Dataset Viewer
index
stringlengths 21
21
| question
stringlengths 87
574
| A
stringlengths 3
221
| B
stringlengths 2
279
| C
stringclasses 1
value | D
stringclasses 1
value | bounding_box
listlengths 1
3
| direction
listlengths 0
3
| answer
stringclasses 4
values | answer_cot
stringlengths 387
4.87k
| answer_name
stringlengths 6
283
| category
stringclasses 12
values | image_url
stringlengths 20
20
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
00002f4ff380c64c_6c30 | Consider the real-world 3D location of the objects. Which object is closer to the camera? | a boat that is old and dirty | a metal fence | null | null | [
{
"bbox_3d": [
-4.5,
0.9,
13.4
],
"label": "a boat that is old and dirty"
},
{
"bbox_3d": [
4.5,
1.4,
19.4
],
"label": "a metal fence"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a boat that is old and dirty and a metal fence. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a boat that is old and dirty is (-4.5, 0.9, 13.4). The 3D location of a metal fence is (4.5, 1.4, 19.4). The L2 distance from the camera to a boat that is old and dirty is 14.19. The L2 distance from the camera to a metal fence is 19.97. The distance to a boat that is old and dirty is smaller. Therefore, the answer is A. a boat that is old and dirty. | A. a boat that is old and dirty. | location_closer_to_camera | 00002f4ff380c64c.jpg |
0000b7e1500c94d7_83e2 | Consider the real-world 3D locations and orientations of the objects. Which side of a black stone monument is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
1.7,
2.2,
5.6
],
"label": "a black stone monument"
}
] | [
{
"front_dir": [
-0.2,
-0.1,
-1
],
"label": "a black stone monument",
"left_dir": [
-1,
0.1,
0.2
]
}
] | A | To solve this problem, we first estimate the 3D location of a black stone monument. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black stone monument, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black stone monument that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a black stone monument is (1.7, 2.2, 5.6). The vector from a black stone monument to camera is hence (-1.7, -2.2, -5.6). The left direction of a black stone monument is (-1.0, 0.1, 0.2). The cosine similarity between the vector pointing to camera and the left direction is 0.05, corresponding to an angle of 87.29 degrees. Thus the angle between the vector pointing to camera and the right direction is 92.71 degrees. The front direction of a black stone monument is (-0.2, -0.1, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.96, corresponding to an angle of 15.31 degrees. Thus the angle between the vector pointing to camera and the back direction is 164.69 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 15.31 degrees. Thus the front side of a black stone monument is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 0000b7e1500c94d7.jpg |
0000bcb094764718_861d | Consider the real-world 3D orientations of the objects. Are a bulletin board with a yellow sign and a green sign and a small computer monitor facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
0.6,
2.3,
4.1
],
"label": "a bulletin board with a yellow sign and a green sign"
},
{
"bbox_3d": [
-1.4,
1.5,
3.7
],
"label": "a small computer monitor"
}
] | [
{
"front_dir": [
-0.1,
-0.2,
-1
],
"label": "a bulletin board with a yellow sign and a green sign",
"left_dir": [
-1,
0,
0.1
]
},
{
"front_dir": [
0.4,
0,
-0.9
],
"label": "a small computer monitor",
"left_dir": [
-0.9,
0.1,
-0.4
]
}
] | A | To solve this problem, we first detect the front directions of a bulletin board with a yellow sign and a green sign and a small computer monitor. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a bulletin board with a yellow sign and a green sign is (-0.1, -0.2, -1.0). The front direction of a small computer monitor is (0.4, 0.0, -0.9). The cosine similarity between the two front directions is 0.86, corresponding to an angle of 30.94. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions. | A. same or similar directions. | multi_object_same_direction | 0000bcb094764718.jpg |
0000eb5027281f2a_6e83 | Consider the real-world 3D locations of the objects. Which is closer to a dark object with a bright light in the background, a man singing into a microphone or a silver instrument? | a man singing into a microphone | a silver instrument | null | null | [
{
"bbox_3d": [
-1.4,
2,
3.2
],
"label": "a dark object with a bright light in the background"
},
{
"bbox_3d": [
0.7,
2.2,
3.3
],
"label": "a man singing into a microphone"
},
{
"bbox_3d": [
0.7,
0.2,
1.5
],
"label": "a silver instrument"
}
] | [] | A | To solve this problem, we first detect the 3D location of a dark object with a bright light in the background, a man singing into a microphone, and a silver instrument. Then we compute the L2 distances between a dark object with a bright light in the background and a man singing into a microphone, and between a dark object with a bright light in the background and a silver instrument. The object that is closer to a dark object with a bright light in the background is the one with a smaller distance. The 3D location of a dark object with a bright light in the background is (-1.4, 2.0, 3.2). The 3D location of a man singing into a microphone is (0.7, 2.2, 3.3). The 3D location of a silver instrument is (0.7, 0.2, 1.5). The L2 distance between a dark object with a bright light in the background and a man singing into a microphone is 2.122002787594441. The L2 distance between a dark object with a bright light in the background and a silver instrument is 3.2015946200067784. Between the two distances, the distance between a dark object with a bright light in the background and a man singing into a microphone is smaller. Therefore, the final answer is A. a man singing into a microphone. | A. a man singing into a microphone. | multi_object_closer_to | 0000eb5027281f2a.jpg |
00017656855f96c7_765e | Consider the real-world 3D locations and orientations of the objects. Which side of a white speaker with a black cover is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-1.6,
3.1,
6.7
],
"label": "a white speaker with a black cover"
}
] | [
{
"front_dir": [
0.9,
-0.1,
0.3
],
"label": "a white speaker with a black cover",
"left_dir": [
0.3,
0,
-0.9
]
}
] | B | To solve this problem, we first estimate the 3D location of a white speaker with a black cover. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white speaker with a black cover, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white speaker with a black cover that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a white speaker with a black cover is (-1.6, 3.1, 6.7). The vector from a white speaker with a black cover to camera is hence (1.6, -3.1, -6.7). The left direction of a white speaker with a black cover is (0.3, -0.0, -0.9). The cosine similarity between the vector pointing to camera and the left direction is 0.91, corresponding to an angle of 24.36 degrees. Thus the angle between the vector pointing to camera and the right direction is 155.64 degrees. The front direction of a white speaker with a black cover is (0.9, -0.1, 0.3). The cosine similarity between the vector pointing to camera and the front direction is -0.05, corresponding to an angle of 92.65 degrees. Thus the angle between the vector pointing to camera and the back direction is 87.35 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 24.36 degrees. Thus the left side of a white speaker with a black cover is facing the camera. Therefore, the final answer is B. left. | B. left. | orientation_viewpoint | 00017656855f96c7.jpg |
00019bc020b24b32_9cba | Consider the real-world 3D locations and orientations of the objects. If I stand at a bulletin board with a hotel sign on it's position facing where it is facing, is a young boy in a red shirt on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
-1.1,
0.3,
2.5
],
"label": "a young boy in a red shirt"
},
{
"bbox_3d": [
0.6,
1.4,
1.8
],
"label": "a bulletin board with a hotel sign on it"
}
] | [
{
"front_dir": [
-0.2,
0,
-1
],
"label": "a bulletin board with a hotel sign on it",
"left_dir": [
-1,
-0.1,
0.2
]
}
] | A | To solve this problem, we first determine the 3D locations of a young boy in a red shirt and a bulletin board with a hotel sign on it. Then we estimate the vector pointing from a bulletin board with a hotel sign on it to a young boy in a red shirt, as well as the left direction of a bulletin board with a hotel sign on it. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a young boy in a red shirt is on the left of a bulletin board with a hotel sign on it. Otherwise, a young boy in a red shirt is behind a bulletin board with a hotel sign on it. The 3D location of a young boy in a red shirt is (-1.1, 0.3, 2.5). The 3D location of a bulletin board with a hotel sign on it is (0.6, 1.4, 1.8). The vector from a bulletin board with a hotel sign on it to a young boy in a red shirt is hence (-1.6, -1.1, 0.7). The left direction of a bulletin board with a hotel sign on it is (-1.0, -0.1, 0.2). The cosine similarity between the vector and the left direction is 0.86, corresponding to an angle of 31.03 degrees. The angle is smaller than 90 degrees, meaning that a young boy in a red shirt is on the left of a bulletin board with a hotel sign on it. Therefore, the final answer is A. on the left. | A. on the left. | orientation_on_the_left | 00019bc020b24b32.jpg |
00019f5540822905_f5eb | Consider the real-world 3D locations and orientations of the objects. If I stand at a long dock's position facing where it is facing, is a tall metal structure on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
55.4,
19.8,
151.5
],
"label": "a tall metal structure"
},
{
"bbox_3d": [
8.7,
8.1,
117.5
],
"label": "a long dock"
}
] | [
{
"front_dir": [
0,
-0.1,
-1
],
"label": "a long dock",
"left_dir": [
-1,
0,
0
]
}
] | B | To solve this problem, we first determine the 3D locations of a tall metal structure and a long dock. Then we estimate the vector pointing from a long dock to a tall metal structure, as well as the left direction of a long dock. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a tall metal structure is on the left of a long dock. Otherwise, a tall metal structure is behind a long dock. The 3D location of a tall metal structure is (55.4, 19.8, 151.5). The 3D location of a long dock is (8.7, 8.1, 117.5). The vector from a long dock to a tall metal structure is hence (46.8, 11.7, 34.0). The left direction of a long dock is (-1.0, 0.0, -0.0). The cosine similarity between the vector and the left direction is -0.78, corresponding to an angle of 141.71 degrees. The angle is smaller than 90 degrees, meaning that a tall metal structure is on the right of a long dock. Therefore, the final answer is B. on the right. | B. on the right. | orientation_on_the_left | 00019f5540822905.jpg |
0002c799b0cd7412_3b49 | Consider the real-world 3D orientations of the objects. Are a black counter with a water bottle on it and a white board with black writing facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
-0.1,
0.4,
2.1
],
"label": "a black counter with a water bottle on it"
},
{
"bbox_3d": [
-1,
0.8,
2.7
],
"label": "a white board with black writing"
}
] | [
{
"front_dir": [
-0.1,
0.1,
-1
],
"label": "a black counter with a water bottle on it",
"left_dir": [
-1,
0.1,
0.1
]
},
{
"front_dir": [
0.4,
0.1,
-0.9
],
"label": "a white board with black writing",
"left_dir": [
-0.9,
0.1,
-0.4
]
}
] | A | To solve this problem, we first detect the front directions of a black counter with a water bottle on it and a white board with black writing. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black counter with a water bottle on it is (-0.1, 0.1, -1.0). The front direction of a white board with black writing is (0.4, 0.1, -0.9). The cosine similarity between the two front directions is 0.89, corresponding to an angle of 27.57. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions. | A. same or similar directions. | multi_object_same_direction | 0002c799b0cd7412.jpg |
0002c799b0cd7412_d658 | Consider the real-world 3D locations and orientations of the objects. Which object is a white board with black writing facing towards, a ventilation fan or the a girl with a ponytail wearing a plaid shirt? | a ventilation fan | a girl with a ponytail wearing a plaid shirt | null | null | [
{
"bbox_3d": [
-1,
0.8,
2.7
],
"label": "a white board with black writing"
},
{
"bbox_3d": [
0.1,
1.4,
2.6
],
"label": "a ventilation fan"
},
{
"bbox_3d": [
0.3,
0.5,
1.6
],
"label": "a girl with a ponytail wearing a plaid shirt"
}
] | [
{
"front_dir": [
0.4,
0.1,
-0.9
],
"label": "a white board with black writing",
"left_dir": [
-0.9,
0.1,
-0.4
]
},
{
"front_dir": [
0,
-0.3,
-0.9
],
"label": "a ventilation fan",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | B | To solve this problem, we first detect the 3D location of a white board with black writing, a ventilation fan, and a girl with a ponytail wearing a plaid shirt. Then we compute the cosine similarities between the front direction of a white board with black writing and the vectors from a white board with black writing to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a white board with black writing is facing towards. The 3D location of a white board with black writing is (-1.0, 0.8, 2.7). The 3D location of a ventilation fan is (0.1, 1.4, 2.6). The 3D location of a girl with a ponytail wearing a plaid shirt is (0.3, 0.5, 1.6). The front direction of a white board with black writing is (0.4, 0.1, -0.9). First we consider if a white board with black writing is facing towards the a ventilation fan. The vector from a white board with black writing to a ventilation fan is (1.0, 0.6, -0.1). The cosine similarity between the front direction and the vector is 0.45, corresponding to an angle of 63.34 degrees. First we consider if a white board with black writing is facing towards the a girl with a ponytail wearing a plaid shirt. The vector from a white board with black writing to a girl with a ponytail wearing a plaid shirt is (1.2, -0.2, -1.0). The cosine similarity between the front direction and the vector is 0.89, corresponding to an angle of 27.40 degrees. We find that the angle between the front direction and a girl with a ponytail wearing a plaid shirt is smaller. Therefore, the final answer is B. a girl with a ponytail wearing a plaid shirt. | B. a girl with a ponytail wearing a plaid shirt. | multi_object_facing | 0002c799b0cd7412.jpg |
0002f9c260441eb0_37ab | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a dishwasher with a glass door and a black and white dishwasher, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
-0.1,
0.4,
0.8
],
"label": "a dishwasher with a glass door"
},
{
"bbox_3d": [
0.2,
0.3,
1
],
"label": "a black and white dishwasher"
}
] | [
{
"front_dir": [
-0.5,
0.1,
-0.9
],
"label": "a dishwasher with a glass door",
"left_dir": [
-0.9,
0,
0.5
]
},
{
"front_dir": [
-0.7,
-0.2,
-0.7
],
"label": "a black and white dishwasher",
"left_dir": [
-0.7,
0.2,
0.7
]
}
] | A | To solve this problem, we first detect the front directions of a dishwasher with a glass door and a black and white dishwasher. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a dishwasher with a glass door is (-0.5, 0.1, -0.9). The front direction of a black and white dishwasher is (-0.7, -0.2, -0.7). The cosine similarity between the two front directions is 0.91, corresponding to an angle of 23.82. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 0002f9c260441eb0.jpg |
0003ff86b3294d34_74be | Consider the real-world 3D locations and orientations of the objects. If I stand at a clock with roman numerals's position facing where it is facing, is an archway in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
-0.9,
6.7,
24.6
],
"label": "an archway"
},
{
"bbox_3d": [
-9.4,
29.8,
46.8
],
"label": "a clock with roman numerals"
}
] | [
{
"front_dir": [
0.2,
-0.9,
-0.4
],
"label": "a clock with roman numerals",
"left_dir": [
-1,
-0.1,
-0.2
]
}
] | A | To solve this problem, we first determine the 3D locations of an archway and a clock with roman numerals. Then we estimate the vector pointing from a clock with roman numerals to an archway, as well as the front direction of a clock with roman numerals. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then an archway is in front of a clock with roman numerals. Otherwise, an archway is behind a clock with roman numerals. The 3D location of an archway is (-0.9, 6.7, 24.6). The 3D location of a clock with roman numerals is (-9.4, 29.8, 46.8). The vector from a clock with roman numerals to an archway is hence (8.5, -23.1, -22.2). The front direction of a clock with roman numerals is (0.2, -0.9, -0.4). The cosine similarity between the vector and the front direction is 0.95, corresponding to an angle of 18.45 degrees. The angle is smaller than 90 degrees, meaning that an archway is in front of a clock with roman numerals. Therefore, the final answer is A. in front of. | A. in front of. | orientation_in_front_of | 0003ff86b3294d34.jpg |
0005dadf2a8e8bef_6245 | Consider the real-world 3D locations of the objects. Are the a flag with red, white and blue stripes and the a painting of a boat on a river next to each other or far away from each other? | next to each other | far away from each other | null | null | [
{
"bbox_3d": [
-0.2,
1.2,
6.6
],
"label": "a flag with red, white and blue stripes"
},
{
"bbox_3d": [
0.1,
0.3,
6.3
],
"label": "a painting of a boat on a river"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a flag with red, white and blue stripes and a painting of a boat on a river. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a flag with red, white and blue stripes is (-0.2, 1.2, 6.6). The 3D location of a painting of a boat on a river is (0.1, 0.3, 6.3). The L2 distance between the two objects is 0.98. The size of the a flag with red, white and blue stripes is roughly 0.22. The size of the a painting of a boat on a river is roughly 2.22. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other. | A. next to each other. | location_next_to | 0005dadf2a8e8bef.jpg |
0005e144616a4198_fc08 | Consider the real-world 3D locations of the objects. Which object has a higher location? | a wooden plank with a brown color and a weathered texture | a green and brown ax | null | null | [
{
"bbox_3d": [
0.2,
0.5,
2.8
],
"label": "a wooden plank with a brown color and a weathered texture"
},
{
"bbox_3d": [
0.5,
1,
3.5
],
"label": "a green and brown ax"
}
] | [] | B | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a wooden plank with a brown color and a weathered texture is 0.7. The 3D height of a green and brown ax is 1.3. The 3D height of a green and brown ax is larger, meaning that the location of a green and brown ax is higher. Therefore, the answer is B. a green and brown ax. | B. a green and brown ax. | height_higher | 0005e144616a4198.jpg |
0006176101d9d341_4688 | Consider the real-world 3D locations and orientations of the objects. If I stand at a red car with a black tarp on the back's position facing where it is facing, is a dusty road in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
-2.3,
0.5,
7.1
],
"label": "a dusty road"
},
{
"bbox_3d": [
1,
1,
7.3
],
"label": "a red car with a black tarp on the back"
}
] | [
{
"front_dir": [
1,
0.1,
0
],
"label": "a red car with a black tarp on the back",
"left_dir": [
0,
0.1,
-1
]
}
] | B | To solve this problem, we first determine the 3D locations of a dusty road and a red car with a black tarp on the back. Then we estimate the vector pointing from a red car with a black tarp on the back to a dusty road, as well as the front direction of a red car with a black tarp on the back. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a dusty road is in front of a red car with a black tarp on the back. Otherwise, a dusty road is behind a red car with a black tarp on the back. The 3D location of a dusty road is (-2.3, 0.5, 7.1). The 3D location of a red car with a black tarp on the back is (1.0, 1.0, 7.3). The vector from a red car with a black tarp on the back to a dusty road is hence (-3.3, -0.5, -0.2). The front direction of a red car with a black tarp on the back is (1.0, 0.1, -0.0). The cosine similarity between the vector and the front direction is -0.99, corresponding to an angle of 173.63 degrees. The angle is smaller than 90 degrees, meaning that a dusty road is behind a red car with a black tarp on the back. Therefore, the final answer is B. behind. | B. behind. | orientation_in_front_of | 0006176101d9d341.jpg |
0006578cb5100265_0797 | Consider the real-world 3D orientations of the objects. Are a boat in the water and a long dock facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
1.2,
0.5,
4.5
],
"label": "a boat in the water"
},
{
"bbox_3d": [
0.4,
0.5,
4.6
],
"label": "a long dock"
}
] | [
{
"front_dir": [
-0.2,
-0.1,
-1
],
"label": "a boat in the water",
"left_dir": [
-1,
0.1,
0.2
]
},
{
"front_dir": [
0,
-0.1,
-1
],
"label": "a long dock",
"left_dir": [
-1,
0.1,
0
]
}
] | A | To solve this problem, we first detect the front directions of a boat in the water and a long dock. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a boat in the water is (-0.2, -0.1, -1.0). The front direction of a long dock is (-0.0, -0.1, -1.0). The cosine similarity between the two front directions is 0.98, corresponding to an angle of 10.16. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions. | A. same or similar directions. | multi_object_same_direction | 0006578cb5100265.jpg |
000731dd242627f2_d4e9 | Consider the real-world 3D locations and orientations of the objects. If I stand at a white cell phone's position facing where it is facing, is a man in a brown uniform on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
-0.3,
0.2,
1.1
],
"label": "a man in a brown uniform"
},
{
"bbox_3d": [
0.4,
0.4,
1.2
],
"label": "a white cell phone"
}
] | [
{
"front_dir": [
-0.3,
-0.2,
-0.9
],
"label": "a white cell phone",
"left_dir": [
-1,
0,
0.3
]
}
] | A | To solve this problem, we first determine the 3D locations of a man in a brown uniform and a white cell phone. Then we estimate the vector pointing from a white cell phone to a man in a brown uniform, as well as the left direction of a white cell phone. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man in a brown uniform is on the left of a white cell phone. Otherwise, a man in a brown uniform is behind a white cell phone. The 3D location of a man in a brown uniform is (-0.3, 0.2, 1.1). The 3D location of a white cell phone is (0.4, 0.4, 1.2). The vector from a white cell phone to a man in a brown uniform is hence (-0.8, -0.3, -0.1). The left direction of a white cell phone is (-1.0, 0.0, 0.3). The cosine similarity between the vector and the left direction is 0.87, corresponding to an angle of 29.69 degrees. The angle is smaller than 90 degrees, meaning that a man in a brown uniform is on the left of a white cell phone. Therefore, the final answer is A. on the left. | A. on the left. | orientation_on_the_left | 000731dd242627f2.jpg |
000780f791b71252_bf67 | Consider the real-world 3D locations of the objects. Are the a red and white bus with a red seat and the a person next to each other or far away from each other? | next to each other | far away from each other | null | null | [
{
"bbox_3d": [
0.2,
1.3,
2.5
],
"label": "a red and white bus with a red seat"
},
{
"bbox_3d": [
0.7,
0.9,
2.5
],
"label": "a person"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a red and white bus with a red seat and a person. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a red and white bus with a red seat is (0.2, 1.3, 2.5). The 3D location of a person is (0.7, 0.9, 2.5). The L2 distance between the two objects is 0.59. The size of the a red and white bus with a red seat is roughly 1.85. The size of the a person is roughly 0.32. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other. | A. next to each other. | location_next_to | 000780f791b71252.jpg |
000780f791b71252_bbc4 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a red and white bus with a red seat and a red and white bus, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.2,
1.3,
2.5
],
"label": "a red and white bus with a red seat"
},
{
"bbox_3d": [
-0.8,
1,
1.5
],
"label": "a red and white bus"
}
] | [
{
"front_dir": [
0.3,
-0.1,
-1
],
"label": "a red and white bus with a red seat",
"left_dir": [
-1,
0,
-0.3
]
},
{
"front_dir": [
0.8,
-0.1,
0.6
],
"label": "a red and white bus",
"left_dir": [
0.6,
0,
-0.8
]
}
] | B | To solve this problem, we first detect the front directions of a red and white bus with a red seat and a red and white bus. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a red and white bus with a red seat is (0.3, -0.1, -1.0). The front direction of a red and white bus is (0.8, -0.1, 0.6). The cosine similarity between the two front directions is -0.29, corresponding to an angle of 106.98. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular. | B. perpendicular. | multi_object_parallel | 000780f791b71252.jpg |
00078d9de5011c3b_f657 | Consider the real-world 3D locations and orientations of the objects. If I stand at a red couch with a white blanket's position facing where it is facing, is a yellow book with a black object on the cover in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
0.1,
1.3,
2.6
],
"label": "a yellow book with a black object on the cover"
},
{
"bbox_3d": [
-0.7,
0.8,
3.5
],
"label": "a red couch with a white blanket"
}
] | [
{
"front_dir": [
0.1,
0.1,
-1
],
"label": "a red couch with a white blanket",
"left_dir": [
-0.9,
0.3,
0
]
}
] | A | To solve this problem, we first determine the 3D locations of a yellow book with a black object on the cover and a red couch with a white blanket. Then we estimate the vector pointing from a red couch with a white blanket to a yellow book with a black object on the cover, as well as the front direction of a red couch with a white blanket. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a yellow book with a black object on the cover is in front of a red couch with a white blanket. Otherwise, a yellow book with a black object on the cover is behind a red couch with a white blanket. The 3D location of a yellow book with a black object on the cover is (0.1, 1.3, 2.6). The 3D location of a red couch with a white blanket is (-0.7, 0.8, 3.5). The vector from a red couch with a white blanket to a yellow book with a black object on the cover is hence (0.8, 0.5, -1.0). The front direction of a red couch with a white blanket is (0.1, 0.1, -1.0). The cosine similarity between the vector and the front direction is 0.79, corresponding to an angle of 37.78 degrees. The angle is smaller than 90 degrees, meaning that a yellow book with a black object on the cover is in front of a red couch with a white blanket. Therefore, the final answer is A. in front of. | A. in front of. | orientation_in_front_of | 00078d9de5011c3b.jpg |
000793c5cff77ab0_2e5e | Consider the real-world 3D locations and orientations of the objects. Which side of a stone building with a window is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-2.4,
2.5,
9.6
],
"label": "a stone building with a window"
}
] | [
{
"front_dir": [
0.2,
-0.1,
-1
],
"label": "a stone building with a window",
"left_dir": [
-1,
0.1,
-0.2
]
}
] | A | To solve this problem, we first estimate the 3D location of a stone building with a window. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a stone building with a window, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a stone building with a window that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a stone building with a window is (-2.4, 2.5, 9.6). The vector from a stone building with a window to camera is hence (2.4, -2.5, -9.6). The left direction of a stone building with a window is (-1.0, 0.1, -0.2). The cosine similarity between the vector pointing to camera and the left direction is -0.05, corresponding to an angle of 92.72 degrees. Thus the angle between the vector pointing to camera and the right direction is 87.28 degrees. The front direction of a stone building with a window is (0.2, -0.1, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.99, corresponding to an angle of 6.99 degrees. Thus the angle between the vector pointing to camera and the back direction is 173.01 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 6.99 degrees. Thus the front side of a stone building with a window is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 000793c5cff77ab0.jpg |
0007c3f55f707547_d1b4 | Consider the real-world 3D locations of the objects. Are the a man with a big smile and the a black hat with a black band next to each other or far away from each other? | next to each other | far away from each other | null | null | [
{
"bbox_3d": [
0.2,
0.3,
1.1
],
"label": "a man with a big smile"
},
{
"bbox_3d": [
0.2,
0.5,
1.2
],
"label": "a black hat with a black band"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a man with a big smile and a black hat with a black band. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a man with a big smile is (0.2, 0.3, 1.1). The 3D location of a black hat with a black band is (0.2, 0.5, 1.2). The L2 distance between the two objects is 0.25. The size of the a man with a big smile is roughly 0.13. The size of the a black hat with a black band is roughly 0.68. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other. | A. next to each other. | location_next_to | 0007c3f55f707547.jpg |
0007c911c52c1324_97ce | Consider the real-world 3D locations and orientations of the objects. Which side of a black camera with a purple fingernail is facing a girl sitting on a rock? | front | left | back | right | [
{
"bbox_3d": [
0.3,
1.3,
2.6
],
"label": "a black camera with a purple fingernail"
},
{
"bbox_3d": [
0.3,
1.4,
2.9
],
"label": "a girl sitting on a rock"
}
] | [
{
"front_dir": [
0,
-0.2,
-1
],
"label": "a black camera with a purple fingernail",
"left_dir": [
-1,
0.1,
0
]
}
] | C | To solve this problem, we first detect the 3D locations of a black camera with a purple fingernail and a girl sitting on a rock. Then we compute the vector pointing from a black camera with a purple fingernail to a girl sitting on a rock. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black camera with a purple fingernail, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black camera with a purple fingernail that is facing a girl sitting on a rock corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a black camera with a purple fingernail is (0.3, 1.3, 2.6). The 3D location of a girl sitting on a rock is (0.3, 1.4, 2.9). The vector from a black camera with a purple fingernail to a girl sitting on a rock is hence (-0.0, 0.1, 0.4). The left direction of a black camera with a purple fingernail is (-1.0, 0.1, 0.0). The cosine similarity between the vector pointing to a girl sitting on a rock and the left direction is 0.06, corresponding to an angle of 86.63 degrees. Thus the angle between the vector pointing to a girl sitting on a rock and the right direction is 93.37 degrees. The front direction of a black camera with a purple fingernail is (-0.0, -0.2, -1.0). The cosine similarity between the vector pointing to a girl sitting on a rock and the front direction is -0.99, corresponding to an angle of 171.46 degrees. Thus the angle between the vector pointing to a girl sitting on a rock and the back direction is 8.54 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 8.54 degrees. Thus the back side of a black camera with a purple fingernail is facing the a girl sitting on a rock. Therefore, the final answer is C. back. | C. back. | multi_object_viewpoint_towards_object | 0007c911c52c1324.jpg |
0007e4ace390f6ee_ce8f | Consider the real-world 3D locations and orientations of the objects. Which object is a tall red and white lighthouse facing towards, a body of water or the a metal railing? | a body of water | a metal railing | null | null | [
{
"bbox_3d": [
-0.8,
-6.7,
49.2
],
"label": "a tall red and white lighthouse"
},
{
"bbox_3d": [
-22.6,
-22.5,
136.4
],
"label": "a body of water"
},
{
"bbox_3d": [
-2.1,
0.9,
10.1
],
"label": "a metal railing"
}
] | [
{
"front_dir": [
-0.5,
0.2,
-0.8
],
"label": "a tall red and white lighthouse",
"left_dir": [
-0.8,
0,
0.5
]
}
] | B | To solve this problem, we first detect the 3D location of a tall red and white lighthouse, a body of water, and a metal railing. Then we compute the cosine similarities between the front direction of a tall red and white lighthouse and the vectors from a tall red and white lighthouse to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a tall red and white lighthouse is facing towards. The 3D location of a tall red and white lighthouse is (-0.8, -6.7, 49.2). The 3D location of a body of water is (-22.6, -22.5, 136.4). The 3D location of a metal railing is (-2.1, 0.9, 10.1). The front direction of a tall red and white lighthouse is (-0.5, 0.2, -0.8). First we consider if a tall red and white lighthouse is facing towards the a body of water. The vector from a tall red and white lighthouse to a body of water is (-21.9, -15.8, 87.1). The cosine similarity between the front direction and the vector is -0.71, corresponding to an angle of 134.99 degrees. First we consider if a tall red and white lighthouse is facing towards the a metal railing. The vector from a tall red and white lighthouse to a metal railing is (-1.3, 7.6, -39.2). The cosine similarity between the front direction and the vector is 0.87, corresponding to an angle of 29.26 degrees. We find that the angle between the front direction and a metal railing is smaller. Therefore, the final answer is B. a metal railing. | B. a metal railing. | multi_object_facing | 0007e4ace390f6ee.jpg |
00084a0cc3a13a3b_18b4 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a black wheelchair and a staircase with a person in a wheelchair at the bottom, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
1.7,
1.7,
38.1
],
"label": "a black wheelchair"
},
{
"bbox_3d": [
0.3,
2.7,
45.7
],
"label": "a staircase with a person in a wheelchair at the bottom"
}
] | [
{
"front_dir": [
0,
0,
-1
],
"label": "a black wheelchair",
"left_dir": [
-1,
0.1,
0
]
},
{
"front_dir": [
0.1,
0,
-1
],
"label": "a staircase with a person in a wheelchair at the bottom",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | A | To solve this problem, we first detect the front directions of a black wheelchair and a staircase with a person in a wheelchair at the bottom. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a black wheelchair is (0.0, 0.0, -1.0). The front direction of a staircase with a person in a wheelchair at the bottom is (0.1, 0.0, -1.0). The cosine similarity between the two front directions is 1.00, corresponding to an angle of 2.40. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 00084a0cc3a13a3b.jpg |
0008712fe7c1e16b_57f6 | Consider the real-world 3D locations of the objects. Which is closer to a white barge with a yellow roof, a black boat with red life preservers or a dock with a ramp? | a black boat with red life preservers | a dock with a ramp | null | null | [
{
"bbox_3d": [
-2.2,
8.6,
79.8
],
"label": "a white barge with a yellow roof"
},
{
"bbox_3d": [
-8.1,
4.5,
56.3
],
"label": "a black boat with red life preservers"
},
{
"bbox_3d": [
-1,
5.3,
67.5
],
"label": "a dock with a ramp"
}
] | [] | B | To solve this problem, we first detect the 3D location of a white barge with a yellow roof, a black boat with red life preservers, and a dock with a ramp. Then we compute the L2 distances between a white barge with a yellow roof and a black boat with red life preservers, and between a white barge with a yellow roof and a dock with a ramp. The object that is closer to a white barge with a yellow roof is the one with a smaller distance. The 3D location of a white barge with a yellow roof is (-2.2, 8.6, 79.8). The 3D location of a black boat with red life preservers is (-8.1, 4.5, 56.3). The 3D location of a dock with a ramp is (-1.0, 5.3, 67.5). The L2 distance between a white barge with a yellow roof and a black boat with red life preservers is 24.57661569024023. The L2 distance between a white barge with a yellow roof and a dock with a ramp is 12.75021364643461. Between the two distances, the distance between a white barge with a yellow roof and a dock with a ramp is smaller. Therefore, the final answer is B. a dock with a ramp. | B. a dock with a ramp. | multi_object_closer_to | 0008712fe7c1e16b.jpg |
000ab7bec71cc50a_a6d3 | Consider the real-world 3D locations of the objects. Are the a piece of bread with butter on it and the a syrup next to each other or far away from each other? | next to each other | far away from each other | null | null | [
{
"bbox_3d": [
-0.2,
0.1,
0.7
],
"label": "a piece of bread with butter on it"
},
{
"bbox_3d": [
0.1,
0.1,
0.7
],
"label": "a syrup"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a piece of bread with butter on it and a syrup. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a piece of bread with butter on it is (-0.2, 0.1, 0.7). The 3D location of a syrup is (0.1, 0.1, 0.7). The L2 distance between the two objects is 0.36. The size of the a piece of bread with butter on it is roughly 0.24. The size of the a syrup is roughly 0.35. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other. | A. next to each other. | location_next_to | 000ab7bec71cc50a.jpg |
000adfe5b817011c_c588 | Consider the real-world 3D locations and orientations of the objects. Which object is a wooden bench with a man sitting on it facing towards, a man sitting on a bench or the a man sitting on a bench? | a man sitting on a bench | a man sitting on a bench | null | null | [
{
"bbox_3d": [
2.2,
0.7,
9.6
],
"label": "a wooden bench with a man sitting on it"
},
{
"bbox_3d": [
1.7,
0.1,
5.9
],
"label": "a man sitting on a bench"
},
{
"bbox_3d": [
0.5,
2.4,
7.7
],
"label": "a man sitting on a bench"
}
] | [
{
"front_dir": [
0.3,
0,
-0.9
],
"label": "a wooden bench with a man sitting on it",
"left_dir": [
-0.9,
0,
-0.3
]
}
] | A | To solve this problem, we first detect the 3D location of a wooden bench with a man sitting on it, a man sitting on a bench, and a man sitting on a bench. Then we compute the cosine similarities between the front direction of a wooden bench with a man sitting on it and the vectors from a wooden bench with a man sitting on it to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a wooden bench with a man sitting on it is facing towards. The 3D location of a wooden bench with a man sitting on it is (2.2, 0.7, 9.6). The 3D location of a man sitting on a bench is (1.7, 0.1, 5.9). The 3D location of a man sitting on a bench is (0.5, 2.4, 7.7). The front direction of a wooden bench with a man sitting on it is (0.3, 0.0, -0.9). First we consider if a wooden bench with a man sitting on it is facing towards the a man sitting on a bench. The vector from a wooden bench with a man sitting on it to a man sitting on a bench is (-0.5, -0.6, -3.8). The cosine similarity between the front direction and the vector is 0.88, corresponding to an angle of 28.45 degrees. First we consider if a wooden bench with a man sitting on it is facing towards the a man sitting on a bench. The vector from a wooden bench with a man sitting on it to a man sitting on a bench is (-1.7, 1.7, -1.9). The cosine similarity between the front direction and the vector is 0.42, corresponding to an angle of 65.19 degrees. We find that the angle between the front direction and a man sitting on a bench is smaller. Therefore, the final answer is A. a man sitting on a bench. | A. a man sitting on a bench. | multi_object_facing | 000adfe5b817011c.jpg |
000b72e1446f8849_1d1a | Consider the real-world 3D orientations of the objects. Are a brown leather couch and a brown leather chair facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
0.9,
1.1,
2.2
],
"label": "a brown leather couch"
},
{
"bbox_3d": [
-0.6,
1.2,
3.6
],
"label": "a brown leather chair"
}
] | [
{
"front_dir": [
-0.6,
0.1,
-0.8
],
"label": "a brown leather couch",
"left_dir": [
-0.8,
-0.1,
0.6
]
},
{
"front_dir": [
0.7,
-0.1,
-0.8
],
"label": "a brown leather chair",
"left_dir": [
-0.8,
0,
-0.7
]
}
] | B | To solve this problem, we first detect the front directions of a brown leather couch and a brown leather chair. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a brown leather couch is (-0.6, 0.1, -0.8). The front direction of a brown leather chair is (0.7, -0.1, -0.8). The cosine similarity between the two front directions is 0.24, corresponding to an angle of 76.22. The angle is large, meaning that the two objects are facing very different directions. Therefore, the final answer is B. very different directions. | B. very different directions. | multi_object_same_direction | 000b72e1446f8849.jpg |
000b72e1446f8849_7d36 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a brown leather couch and a brown leather chair, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.9,
1.1,
2.2
],
"label": "a brown leather couch"
},
{
"bbox_3d": [
-0.6,
1.2,
3.6
],
"label": "a brown leather chair"
}
] | [
{
"front_dir": [
-0.6,
0.1,
-0.8
],
"label": "a brown leather couch",
"left_dir": [
-0.8,
-0.1,
0.6
]
},
{
"front_dir": [
0.7,
-0.1,
-0.8
],
"label": "a brown leather chair",
"left_dir": [
-0.8,
0,
-0.7
]
}
] | B | To solve this problem, we first detect the front directions of a brown leather couch and a brown leather chair. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a brown leather couch is (-0.6, 0.1, -0.8). The front direction of a brown leather chair is (0.7, -0.1, -0.8). The cosine similarity between the two front directions is 0.24, corresponding to an angle of 76.22. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular. | B. perpendicular. | multi_object_parallel | 000b72e1446f8849.jpg |
000c6dd7ff3e2998_d853 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a wooden dock with a rotten post and a large grey battleship, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
1.2,
0.9,
11.1
],
"label": "a wooden dock with a rotten post"
},
{
"bbox_3d": [
1.1,
18.2,
143.7
],
"label": "a large grey battleship"
}
] | [
{
"front_dir": [
0,
0,
-1
],
"label": "a wooden dock with a rotten post",
"left_dir": [
-1,
-0.1,
0
]
},
{
"front_dir": [
-0.1,
0.2,
1
],
"label": "a large grey battleship",
"left_dir": [
1,
0,
0.1
]
}
] | A | To solve this problem, we first detect the front directions of a wooden dock with a rotten post and a large grey battleship. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a wooden dock with a rotten post is (-0.0, -0.0, -1.0). The front direction of a large grey battleship is (-0.1, 0.2, 1.0). The cosine similarity between the two front directions is -0.98, corresponding to an angle of 169.70. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 000c6dd7ff3e2998.jpg |
000c7072d83a9090_0b7a | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a black and silver dresser with a mirror on top and a white and black armchair, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.6,
0.5,
1.3
],
"label": "a black and silver dresser with a mirror on top"
},
{
"bbox_3d": [
-0.2,
0.6,
1.8
],
"label": "a white and black armchair"
}
] | [
{
"front_dir": [
-1,
0.2,
0
],
"label": "a black and silver dresser with a mirror on top",
"left_dir": [
0,
-0.4,
0.9
]
},
{
"front_dir": [
-0.4,
0.1,
-0.9
],
"label": "a white and black armchair",
"left_dir": [
-0.9,
0,
0.4
]
}
] | B | To solve this problem, we first detect the front directions of a black and silver dresser with a mirror on top and a white and black armchair. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a black and silver dresser with a mirror on top is (-1.0, 0.2, 0.0). The front direction of a white and black armchair is (-0.4, 0.1, -0.9). The cosine similarity between the two front directions is 0.38, corresponding to an angle of 67.50. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular. | B. perpendicular. | multi_object_parallel | 000c7072d83a9090.jpg |
000c984366d5aa17_d576 | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a boat in the water | a dock with a boat | null | null | [
{
"bbox_3d": [
-2.3,
1.1,
54.4
],
"label": "a boat in the water"
},
{
"bbox_3d": [
-12.3,
4.4,
126.6
],
"label": "a dock with a boat"
}
] | [] | B | To solve this problem, we first estimate the 3D locations of a boat in the water and a dock with a boat. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a boat in the water is (-2.3, 1.1, 54.4). The 3D location of a dock with a boat is (-12.3, 4.4, 126.6). The L2 distance from the camera to a boat in the water is 54.49. The L2 distance from the camera to a dock with a boat is 127.26. The distance to a dock with a boat is larger. Therefore, the answer is B. a dock with a boat. | B. a dock with a boat. | location_closer_to_camera | 000c984366d5aa17.jpg |
000d05b39cf0d5c6_0b78 | Consider the real-world 3D locations and orientations of the objects. If I stand at a large vehicle with a black door's position facing where it is facing, is a white basketball hoop in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
-3,
2.5,
7.5
],
"label": "a white basketball hoop"
},
{
"bbox_3d": [
-1.6,
0.8,
5.2
],
"label": "a large vehicle with a black door"
}
] | [
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a large vehicle with a black door",
"left_dir": [
-0.9,
-0.1,
-0.4
]
}
] | B | To solve this problem, we first determine the 3D locations of a white basketball hoop and a large vehicle with a black door. Then we estimate the vector pointing from a large vehicle with a black door to a white basketball hoop, as well as the front direction of a large vehicle with a black door. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a white basketball hoop is in front of a large vehicle with a black door. Otherwise, a white basketball hoop is behind a large vehicle with a black door. The 3D location of a white basketball hoop is (-3.0, 2.5, 7.5). The 3D location of a large vehicle with a black door is (-1.6, 0.8, 5.2). The vector from a large vehicle with a black door to a white basketball hoop is hence (-1.4, 1.7, 2.3). The front direction of a large vehicle with a black door is (0.4, -0.1, -0.9). The cosine similarity between the vector and the front direction is -0.89, corresponding to an angle of 152.93 degrees. The angle is smaller than 90 degrees, meaning that a white basketball hoop is behind a large vehicle with a black door. Therefore, the final answer is B. behind. | B. behind. | orientation_in_front_of | 000d05b39cf0d5c6.jpg |
000e70d6b9f25c3d_c947 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a fireplace with a black metal grate and a brick wall and a white couch with a brown pillow, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.2,
0.6,
6.3
],
"label": "a fireplace with a black metal grate and a brick wall"
},
{
"bbox_3d": [
-1.2,
0.5,
4.1
],
"label": "a white couch with a brown pillow"
}
] | [
{
"front_dir": [
-0.1,
0.1,
-1
],
"label": "a fireplace with a black metal grate and a brick wall",
"left_dir": [
-1,
0,
0.1
]
},
{
"front_dir": [
0.2,
0,
-1
],
"label": "a white couch with a brown pillow",
"left_dir": [
-1,
-0.1,
-0.2
]
}
] | A | To solve this problem, we first detect the front directions of a fireplace with a black metal grate and a brick wall and a white couch with a brown pillow. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a fireplace with a black metal grate and a brick wall is (-0.1, 0.1, -1.0). The front direction of a white couch with a brown pillow is (0.2, 0.0, -1.0). The cosine similarity between the two front directions is 0.94, corresponding to an angle of 19.79. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 000e70d6b9f25c3d.jpg |
000efb7df8aa950c_c64c | Consider the real-world 3D locations of the objects. Is a wood wall directly above a woman sitting on a green couch? | yes | no | null | null | [
{
"bbox_3d": [
0,
1.7,
5
],
"label": "a wood wall"
},
{
"bbox_3d": [
0.6,
1.1,
4.7
],
"label": "a woman sitting on a green couch"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a wood wall and a woman sitting on a green couch. Then we compute the vector pointing from a woman sitting on a green couch to a wood wall, as well as the up direction of a woman sitting on a green couch. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a wood wall is directly above a woman sitting on a green couch. Otherwise, then a wood wall is not directly above a woman sitting on a green couch. The 3D location of a wood wall is (-0.0, 1.7, 5.0). The 3D location of a woman sitting on a green couch is (0.6, 1.1, 4.7). The vector from a woman sitting on a green couch to a wood wall is hence (-0.7, 0.6, 0.3). The up direction of a woman sitting on a green couch is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.61, corresponding to an angle of 52 degrees. The angle between the vector and the up direction is large, meaning that a wood wall is not directly above a woman sitting on a green couch. Therefore, the answer is B. no. | B. no. | location_above | 000efb7df8aa950c.jpg |
000f186681cac78d_a956 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a bulletin board with a sign on it and a white and black machine with a green sticker on it, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.1,
2,
3.9
],
"label": "a bulletin board with a sign on it"
},
{
"bbox_3d": [
0,
0.5,
1.4
],
"label": "a white and black machine with a green sticker on it"
}
] | [
{
"front_dir": [
0.1,
-0.3,
-1
],
"label": "a bulletin board with a sign on it",
"left_dir": [
-1,
0,
-0.1
]
},
{
"front_dir": [
0.1,
0,
-1
],
"label": "a white and black machine with a green sticker on it",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | A | To solve this problem, we first detect the front directions of a bulletin board with a sign on it and a white and black machine with a green sticker on it. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a bulletin board with a sign on it is (0.1, -0.3, -1.0). The front direction of a white and black machine with a green sticker on it is (0.1, 0.0, -1.0). The cosine similarity between the two front directions is 0.96, corresponding to an angle of 15.84. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 000f186681cac78d.jpg |
00113a6dc6c71b5e_2671 | Consider the real-world 3D locations and orientations of the objects. If I stand at a white laptop computer's position facing where it is facing, is a glass of wine with a dark red liquid on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
-0.3,
0.7,
1.3
],
"label": "a glass of wine with a dark red liquid"
},
{
"bbox_3d": [
-0.4,
0.8,
1
],
"label": "a white laptop computer"
}
] | [
{
"front_dir": [
0.7,
-0.5,
-0.5
],
"label": "a white laptop computer",
"left_dir": [
-0.7,
-0.2,
-0.7
]
}
] | B | To solve this problem, we first determine the 3D locations of a glass of wine with a dark red liquid and a white laptop computer. Then we estimate the vector pointing from a white laptop computer to a glass of wine with a dark red liquid, as well as the left direction of a white laptop computer. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a glass of wine with a dark red liquid is on the left of a white laptop computer. Otherwise, a glass of wine with a dark red liquid is behind a white laptop computer. The 3D location of a glass of wine with a dark red liquid is (-0.3, 0.7, 1.3). The 3D location of a white laptop computer is (-0.4, 0.8, 1.0). The vector from a white laptop computer to a glass of wine with a dark red liquid is hence (0.0, -0.1, 0.4). The left direction of a white laptop computer is (-0.7, -0.2, -0.7). The cosine similarity between the vector and the left direction is -0.71, corresponding to an angle of 135.51 degrees. The angle is smaller than 90 degrees, meaning that a glass of wine with a dark red liquid is on the right of a white laptop computer. Therefore, the final answer is B. on the right. | B. on the right. | orientation_on_the_left | 00113a6dc6c71b5e.jpg |
001156eb13f37194_1f7a | Consider the real-world 3D locations and orientations of the objects. Which object is a white car with a license plate facing towards, a black and white license plate or the a garage with a white car? | a black and white license plate | a garage with a white car | null | null | [
{
"bbox_3d": [
0.4,
1,
2.5
],
"label": "a white car with a license plate"
},
{
"bbox_3d": [
-0.4,
0.5,
1.7
],
"label": "a black and white license plate"
},
{
"bbox_3d": [
-1.4,
1.5,
5
],
"label": "a garage with a white car"
}
] | [
{
"front_dir": [
-0.5,
0,
-0.8
],
"label": "a white car with a license plate",
"left_dir": [
-0.8,
0.1,
0.5
]
}
] | A | To solve this problem, we first detect the 3D location of a white car with a license plate, a black and white license plate, and a garage with a white car. Then we compute the cosine similarities between the front direction of a white car with a license plate and the vectors from a white car with a license plate to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a white car with a license plate is facing towards. The 3D location of a white car with a license plate is (0.4, 1.0, 2.5). The 3D location of a black and white license plate is (-0.4, 0.5, 1.7). The 3D location of a garage with a white car is (-1.4, 1.5, 5.0). The front direction of a white car with a license plate is (-0.5, -0.0, -0.8). First we consider if a white car with a license plate is facing towards the a black and white license plate. The vector from a white car with a license plate to a black and white license plate is (-0.8, -0.4, -0.8). The cosine similarity between the front direction and the vector is 0.92, corresponding to an angle of 23.69 degrees. First we consider if a white car with a license plate is facing towards the a garage with a white car. The vector from a white car with a license plate to a garage with a white car is (-1.9, 0.6, 2.5). The cosine similarity between the front direction and the vector is -0.36, corresponding to an angle of 111.35 degrees. We find that the angle between the front direction and a black and white license plate is smaller. Therefore, the final answer is A. a black and white license plate. | A. a black and white license plate. | multi_object_facing | 001156eb13f37194.jpg |
0011d04b9979b065_d58b | Consider the real-world 3D locations of the objects. Which object has a lower location? | a fish tail | a green sea horse | null | null | [
{
"bbox_3d": [
0.5,
3.4,
0.7
],
"label": "a fish tail"
},
{
"bbox_3d": [
0,
1.3,
2.2
],
"label": "a green sea horse"
}
] | [] | A | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a fish tail is 3.8. The 3D height of a green sea horse is 4.8. The 3D height of a green sea horse is larger, meaning that the location of a green sea horse is higher. In other words, the location of a fish tail is lower. Therefore, the answer is A. a fish tail. | A. a fish tail | height_higher | 0011d04b9979b065.jpg |
0011f57bd95eb8ce_ab60 | Consider the real-world 3D location of the objects. Which object is closer to the camera? | a book with a black cover | a red mountain bike with a silver frame | null | null | [
{
"bbox_3d": [
0.4,
1.1,
1.6
],
"label": "a book with a black cover"
},
{
"bbox_3d": [
0,
0.8,
1.2
],
"label": "a red mountain bike with a silver frame"
}
] | [] | B | To solve this problem, we first estimate the 3D locations of a book with a black cover and a red mountain bike with a silver frame. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a book with a black cover is (0.4, 1.1, 1.6). The 3D location of a red mountain bike with a silver frame is (0.0, 0.8, 1.2). The L2 distance from the camera to a book with a black cover is 2.04. The L2 distance from the camera to a red mountain bike with a silver frame is 1.46. The distance to a red mountain bike with a silver frame is smaller. Therefore, the answer is B. a red mountain bike with a silver frame. | B. a red mountain bike with a silver frame. | location_closer_to_camera | 0011f57bd95eb8ce.jpg |
0011fe8903af9b19_8bf2 | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a woman in a long dress sitting on a couch | a black and white photo of a person's feet | null | null | [
{
"bbox_3d": [
0.2,
0.8,
2.6
],
"label": "a woman in a long dress sitting on a couch"
},
{
"bbox_3d": [
0.1,
0.2,
2.1
],
"label": "a black and white photo of a person's feet"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a woman in a long dress sitting on a couch and a black and white photo of a person's feet. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a woman in a long dress sitting on a couch is (0.2, 0.8, 2.6). The 3D location of a black and white photo of a person's feet is (0.1, 0.2, 2.1). The L2 distance from the camera to a woman in a long dress sitting on a couch is 2.70. The L2 distance from the camera to a black and white photo of a person's feet is 2.15. The distance to a woman in a long dress sitting on a couch is larger. Therefore, the answer is A. a woman in a long dress sitting on a couch. | A. a woman in a long dress sitting on a couch. | location_closer_to_camera | 0011fe8903af9b19.jpg |
001334773bd8d5c8_2238 | Consider the real-world 3D locations and orientations of the objects. Which side of a car with a black roof is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
1.4,
7.1,
79.5
],
"label": "a car with a black roof"
}
] | [
{
"front_dir": [
0.1,
-0.2,
-1
],
"label": "a car with a black roof",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | A | To solve this problem, we first estimate the 3D location of a car with a black roof. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a car with a black roof, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a car with a black roof that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a car with a black roof is (1.4, 7.1, 79.5). The vector from a car with a black roof to camera is hence (-1.4, -7.1, -79.5). The left direction of a car with a black roof is (-1.0, 0.1, -0.1). The cosine similarity between the vector pointing to camera and the left direction is 0.08, corresponding to an angle of 85.55 degrees. Thus the angle between the vector pointing to camera and the right direction is 94.45 degrees. The front direction of a car with a black roof is (0.1, -0.2, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.99, corresponding to an angle of 5.85 degrees. Thus the angle between the vector pointing to camera and the back direction is 174.15 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.85 degrees. Thus the front side of a car with a black roof is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 001334773bd8d5c8.jpg |
001334773bd8d5c8_9778 | Consider the real-world 3D locations and orientations of the objects. Which object is a car with a black roof facing towards, a black and white photo of a street or the a black and white photo of a road? | a black and white photo of a street | a black and white photo of a road | null | null | [
{
"bbox_3d": [
1.4,
7.1,
79.5
],
"label": "a car with a black roof"
},
{
"bbox_3d": [
5.6,
13.3,
76.1
],
"label": "a black and white photo of a street"
},
{
"bbox_3d": [
1,
1.3,
22.2
],
"label": "a black and white photo of a road"
}
] | [
{
"front_dir": [
0.1,
-0.2,
-1
],
"label": "a car with a black roof",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | B | To solve this problem, we first detect the 3D location of a car with a black roof, a black and white photo of a street, and a black and white photo of a road. Then we compute the cosine similarities between the front direction of a car with a black roof and the vectors from a car with a black roof to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a car with a black roof is facing towards. The 3D location of a car with a black roof is (1.4, 7.1, 79.5). The 3D location of a black and white photo of a street is (5.6, 13.3, 76.1). The 3D location of a black and white photo of a road is (1.0, 1.3, 22.2). The front direction of a car with a black roof is (0.1, -0.2, -1.0). First we consider if a car with a black roof is facing towards the a black and white photo of a street. The vector from a car with a black roof to a black and white photo of a street is (4.1, 6.2, -3.4). The cosine similarity between the front direction and the vector is 0.31, corresponding to an angle of 71.86 degrees. First we consider if a car with a black roof is facing towards the a black and white photo of a road. The vector from a car with a black roof to a black and white photo of a road is (-0.4, -5.8, -57.3). The cosine similarity between the front direction and the vector is 1.00, corresponding to an angle of 4.96 degrees. We find that the angle between the front direction and a black and white photo of a road is smaller. Therefore, the final answer is B. a black and white photo of a road. | B. a black and white photo of a road. | multi_object_facing | 001334773bd8d5c8.jpg |
00134148c2ab3db5_271f | Consider the real-world 3D locations of the objects. Which is closer to a white table with a green cloth on it, a laptop with a white screen or a convention with a green and white tablecloth? | a laptop with a white screen | a convention with a green and white tablecloth | null | null | [
{
"bbox_3d": [
0.3,
0.6,
2
],
"label": "a white table with a green cloth on it"
},
{
"bbox_3d": [
1.8,
1,
5.2
],
"label": "a laptop with a white screen"
},
{
"bbox_3d": [
2.7,
3.2,
6.9
],
"label": "a convention with a green and white tablecloth"
}
] | [] | A | To solve this problem, we first detect the 3D location of a white table with a green cloth on it, a laptop with a white screen, and a convention with a green and white tablecloth. Then we compute the L2 distances between a white table with a green cloth on it and a laptop with a white screen, and between a white table with a green cloth on it and a convention with a green and white tablecloth. The object that is closer to a white table with a green cloth on it is the one with a smaller distance. The 3D location of a white table with a green cloth on it is (0.3, 0.6, 2.0). The 3D location of a laptop with a white screen is (1.8, 1.0, 5.2). The 3D location of a convention with a green and white tablecloth is (2.7, 3.2, 6.9). The L2 distance between a white table with a green cloth on it and a laptop with a white screen is 3.542799601563872. The L2 distance between a white table with a green cloth on it and a convention with a green and white tablecloth is 6.006793476484585. Between the two distances, the distance between a white table with a green cloth on it and a laptop with a white screen is smaller. Therefore, the final answer is A. a laptop with a white screen. | A. a laptop with a white screen. | multi_object_closer_to | 00134148c2ab3db5.jpg |
00136508759eb46c_aa52 | Consider the real-world 3D locations of the objects. Which is closer to a grave with flowers, a concrete slab with a hole in the middle or a headstone? | a concrete slab with a hole in the middle | a headstone | null | null | [
{
"bbox_3d": [
-0.6,
0,
2.3
],
"label": "a grave with flowers"
},
{
"bbox_3d": [
-1,
0.1,
2.5
],
"label": "a concrete slab with a hole in the middle"
},
{
"bbox_3d": [
0,
0.3,
2.4
],
"label": "a headstone"
}
] | [] | A | To solve this problem, we first detect the 3D location of a grave with flowers, a concrete slab with a hole in the middle, and a headstone. Then we compute the L2 distances between a grave with flowers and a concrete slab with a hole in the middle, and between a grave with flowers and a headstone. The object that is closer to a grave with flowers is the one with a smaller distance. The 3D location of a grave with flowers is (-0.6, 0.0, 2.3). The 3D location of a concrete slab with a hole in the middle is (-1.0, 0.1, 2.5). The 3D location of a headstone is (0.0, 0.3, 2.4). The L2 distance between a grave with flowers and a concrete slab with a hole in the middle is 0.5070132367202709. The L2 distance between a grave with flowers and a headstone is 0.6795382385310884. Between the two distances, the distance between a grave with flowers and a concrete slab with a hole in the middle is smaller. Therefore, the final answer is A. a concrete slab with a hole in the middle. | A. a concrete slab with a hole in the middle. | multi_object_closer_to | 00136508759eb46c.jpg |
0013cf20040cbc04_79da | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a silver car and a green and red tow truck, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
-0.8,
0.7,
1.8
],
"label": "a silver car"
},
{
"bbox_3d": [
-0.2,
0.9,
3.2
],
"label": "a green and red tow truck"
}
] | [
{
"front_dir": [
0.8,
0.1,
-0.6
],
"label": "a silver car",
"left_dir": [
-0.6,
0,
-0.8
]
},
{
"front_dir": [
0.7,
0.1,
-0.7
],
"label": "a green and red tow truck",
"left_dir": [
-0.7,
-0.1,
-0.7
]
}
] | A | To solve this problem, we first detect the front directions of a silver car and a green and red tow truck. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a silver car is (0.8, 0.1, -0.6). The front direction of a green and red tow truck is (0.7, 0.1, -0.7). The cosine similarity between the two front directions is 0.98, corresponding to an angle of 11.30. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 0013cf20040cbc04.jpg |
001447f949ca0552_f473 | Consider the real-world 3D locations and orientations of the objects. Which side of a group of people on a porch is facing a woman with a baby? | front | left | back | right | [
{
"bbox_3d": [
-0.2,
1.4,
4.3
],
"label": "a group of people on a porch"
},
{
"bbox_3d": [
0.2,
0.7,
2
],
"label": "a woman with a baby"
}
] | [
{
"front_dir": [
0.1,
-0.2,
-1
],
"label": "a group of people on a porch",
"left_dir": [
-1,
0,
-0.1
]
}
] | A | To solve this problem, we first detect the 3D locations of a group of people on a porch and a woman with a baby. Then we compute the vector pointing from a group of people on a porch to a woman with a baby. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a group of people on a porch, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a group of people on a porch that is facing a woman with a baby corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a group of people on a porch is (-0.2, 1.4, 4.3). The 3D location of a woman with a baby is (0.2, 0.7, 2.0). The vector from a group of people on a porch to a woman with a baby is hence (0.3, -0.8, -2.4). The left direction of a group of people on a porch is (-1.0, -0.0, -0.1). The cosine similarity between the vector pointing to a woman with a baby and the left direction is -0.00, corresponding to an angle of 90.01 degrees. Thus the angle between the vector pointing to a woman with a baby and the right direction is 89.99 degrees. The front direction of a group of people on a porch is (0.1, -0.2, -1.0). The cosine similarity between the vector pointing to a woman with a baby and the front direction is 1.00, corresponding to an angle of 5.50 degrees. Thus the angle between the vector pointing to a woman with a baby and the back direction is 174.50 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.50 degrees. Thus the front side of a group of people on a porch is facing the a woman with a baby. Therefore, the final answer is A. front. | A. front. | multi_object_viewpoint_towards_object | 001447f949ca0552.jpg |
001493fd26374606_175d | Consider the real-world 3D locations of the objects. Which object has a higher location? | a black and white photo | a white fence with a black mesh | null | null | [
{
"bbox_3d": [
-0.9,
0.6,
12.6
],
"label": "a black and white photo"
},
{
"bbox_3d": [
1.2,
1.2,
7.4
],
"label": "a white fence with a black mesh"
}
] | [] | A | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a black and white photo is 2.2. The 3D height of a white fence with a black mesh is 1.5. The 3D height of a black and white photo is larger, meaning that the location of a black and white photo is higher. Therefore, the answer is A. a white fence with a black mesh. | A. a white fence with a black mesh. | height_higher | 001493fd26374606.jpg |
0014c2488ff388be_5d21 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a man in black shirt driving and a red sports car with a license plate, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.6,
1.6,
5.5
],
"label": "a man in black shirt driving"
},
{
"bbox_3d": [
0.9,
1.3,
5
],
"label": "a red sports car with a license plate"
}
] | [
{
"front_dir": [
0.3,
-0.1,
-1
],
"label": "a man in black shirt driving",
"left_dir": [
-1,
-0.1,
-0.3
]
},
{
"front_dir": [
0.4,
0.1,
-0.9
],
"label": "a red sports car with a license plate",
"left_dir": [
-0.9,
0,
-0.4
]
}
] | A | To solve this problem, we first detect the front directions of a man in black shirt driving and a red sports car with a license plate. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a man in black shirt driving is (0.3, -0.1, -1.0). The front direction of a red sports car with a license plate is (0.4, 0.1, -0.9). The cosine similarity between the two front directions is 0.97, corresponding to an angle of 12.95. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 0014c2488ff388be.jpg |
001547a768924358_8af4 | Consider the real-world 3D location of the objects. Which object is closer to the camera? | a gold picture frame with a red cloth | a woman wearing a red and white christmas hat | null | null | [
{
"bbox_3d": [
0.2,
1.6,
1.8
],
"label": "a gold picture frame with a red cloth"
},
{
"bbox_3d": [
0.1,
1.2,
1.1
],
"label": "a woman wearing a red and white christmas hat"
}
] | [] | B | To solve this problem, we first estimate the 3D locations of a gold picture frame with a red cloth and a woman wearing a red and white christmas hat. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a gold picture frame with a red cloth is (0.2, 1.6, 1.8). The 3D location of a woman wearing a red and white christmas hat is (0.1, 1.2, 1.1). The L2 distance from the camera to a gold picture frame with a red cloth is 2.45. The L2 distance from the camera to a woman wearing a red and white christmas hat is 1.63. The distance to a woman wearing a red and white christmas hat is smaller. Therefore, the answer is B. a woman wearing a red and white christmas hat. | B. a woman wearing a red and white christmas hat. | location_closer_to_camera | 001547a768924358.jpg |
001627bf35373371_c3bb | Consider the real-world 3D orientations of the objects. Are a black seat belt and a car is parked on the street facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
0,
0.3,
0.1
],
"label": "a black seat belt"
},
{
"bbox_3d": [
1.8,
-2.2,
3.1
],
"label": "a car is parked on the street"
}
] | [
{
"front_dir": [
1,
-0.1,
0.1
],
"label": "a black seat belt",
"left_dir": [
0.1,
1,
-0.2
]
},
{
"front_dir": [
-0.3,
0.7,
-0.7
],
"label": "a car is parked on the street",
"left_dir": [
-0.9,
-0.4,
0
]
}
] | B | To solve this problem, we first detect the front directions of a black seat belt and a car is parked on the street. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black seat belt is (1.0, -0.1, 0.1). The front direction of a car is parked on the street is (-0.3, 0.7, -0.7). The cosine similarity between the two front directions is -0.40, corresponding to an angle of 113.74. The angle is large, meaning that the two objects are facing very different directions. Therefore, the final answer is B. very different directions. | B. very different directions. | multi_object_same_direction | 001627bf35373371.jpg |
001679a19bb6fd3f_b69a | Consider the real-world 3D locations and orientations of the objects. If I stand at a black car's position facing where it is facing, is a tree with green leaves on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
0.2,
0.9,
10.7
],
"label": "a tree with green leaves"
},
{
"bbox_3d": [
4.3,
0.8,
11.5
],
"label": "a black car"
}
] | [
{
"front_dir": [
0.3,
0,
1
],
"label": "a black car",
"left_dir": [
1,
-0.1,
-0.3
]
}
] | B | To solve this problem, we first determine the 3D locations of a tree with green leaves and a black car. Then we estimate the vector pointing from a black car to a tree with green leaves, as well as the left direction of a black car. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a tree with green leaves is on the left of a black car. Otherwise, a tree with green leaves is behind a black car. The 3D location of a tree with green leaves is (0.2, 0.9, 10.7). The 3D location of a black car is (4.3, 0.8, 11.5). The vector from a black car to a tree with green leaves is hence (-4.1, 0.2, -0.8). The left direction of a black car is (1.0, -0.1, -0.3). The cosine similarity between the vector and the left direction is -0.89, corresponding to an angle of 152.27 degrees. The angle is smaller than 90 degrees, meaning that a tree with green leaves is on the right of a black car. Therefore, the final answer is B. on the right. | B. on the right. | orientation_on_the_left | 001679a19bb6fd3f.jpg |
00173788cafe4c7d_9c4a | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a white and black instrument | a group of men playing guitars | null | null | [
{
"bbox_3d": [
1,
1.5,
5.2
],
"label": "a white and black instrument"
},
{
"bbox_3d": [
-1.1,
0.6,
4.1
],
"label": "a group of men playing guitars"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a white and black instrument and a group of men playing guitars. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a white and black instrument is (1.0, 1.5, 5.2). The 3D location of a group of men playing guitars is (-1.1, 0.6, 4.1). The L2 distance from the camera to a white and black instrument is 5.49. The L2 distance from the camera to a group of men playing guitars is 4.31. The distance to a white and black instrument is larger. Therefore, the answer is A. a white and black instrument. | A. a white and black instrument. | location_closer_to_camera | 00173788cafe4c7d.jpg |
0017633be1cc83c0_2d9b | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a yellow wire | a hand holding a pair of scissors | null | null | [
{
"bbox_3d": [
0.5,
0,
1
],
"label": "a yellow wire"
},
{
"bbox_3d": [
-0.2,
0.4,
0.4
],
"label": "a hand holding a pair of scissors"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a yellow wire and a hand holding a pair of scissors. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a yellow wire is (0.5, 0.0, 1.0). The 3D location of a hand holding a pair of scissors is (-0.2, 0.4, 0.4). The L2 distance from the camera to a yellow wire is 1.13. The L2 distance from the camera to a hand holding a pair of scissors is 0.56. The distance to a yellow wire is larger. Therefore, the answer is A. a yellow wire. | A. a yellow wire. | location_closer_to_camera | 0017633be1cc83c0.jpg |
0017a1a8af3f78a7_7b5e | Consider the real-world 3D location of the objects. Which object is closer to the camera? | a small yellow car is driving down a curvy road | a man wearing blue jeans | null | null | [
{
"bbox_3d": [
0,
0.1,
6.9
],
"label": "a small yellow car is driving down a curvy road"
},
{
"bbox_3d": [
-2.3,
2.1,
15
],
"label": "a man wearing blue jeans"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a small yellow car is driving down a curvy road and a man wearing blue jeans. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a small yellow car is driving down a curvy road is (0.0, 0.1, 6.9). The 3D location of a man wearing blue jeans is (-2.3, 2.1, 15.0). The L2 distance from the camera to a small yellow car is driving down a curvy road is 6.88. The L2 distance from the camera to a man wearing blue jeans is 15.31. The distance to a small yellow car is driving down a curvy road is smaller. Therefore, the answer is A. a small yellow car is driving down a curvy road. | A. a small yellow car is driving down a curvy road. | location_closer_to_camera | 0017a1a8af3f78a7.jpg |
0017a7c9a3027127_41f8 | Consider the real-world 3D locations of the objects. Which object has a lower location? | a tall bookcase | a black and white tie | null | null | [
{
"bbox_3d": [
-1.1,
2,
5.3
],
"label": "a tall bookcase"
},
{
"bbox_3d": [
0.1,
1.2,
2.5
],
"label": "a black and white tie"
}
] | [] | B | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a tall bookcase is 5.6. The 3D height of a black and white tie is 1.6. The 3D height of a tall bookcase is larger, meaning that the location of a tall bookcase is higher. In other words, the location of a black and white tie is lower. Therefore, the answer is B. a tall bookcase. | B. a tall bookcase | height_higher | 0017a7c9a3027127.jpg |
0017abcdf3a66f23_6c84 | Consider the real-world 3D locations of the objects. Which is closer to a person standing on a stairway, a castle with a stone wall or a black sweatshirt? | a castle with a stone wall | a black sweatshirt | null | null | [
{
"bbox_3d": [
-4.8,
2.2,
12.1
],
"label": "a person standing on a stairway"
},
{
"bbox_3d": [
3.3,
3.4,
26.3
],
"label": "a castle with a stone wall"
},
{
"bbox_3d": [
-0.6,
5.1,
2.8
],
"label": "a black sweatshirt"
}
] | [] | B | To solve this problem, we first detect the 3D location of a person standing on a stairway, a castle with a stone wall, and a black sweatshirt. Then we compute the L2 distances between a person standing on a stairway and a castle with a stone wall, and between a person standing on a stairway and a black sweatshirt. The object that is closer to a person standing on a stairway is the one with a smaller distance. The 3D location of a person standing on a stairway is (-4.8, 2.2, 12.1). The 3D location of a castle with a stone wall is (3.3, 3.4, 26.3). The 3D location of a black sweatshirt is (-0.6, 5.1, 2.8). The L2 distance between a person standing on a stairway and a castle with a stone wall is 16.356520637415652. The L2 distance between a person standing on a stairway and a black sweatshirt is 10.631574461816104. Between the two distances, the distance between a person standing on a stairway and a black sweatshirt is smaller. Therefore, the final answer is B. a black sweatshirt. | B. a black sweatshirt. | multi_object_closer_to | 0017abcdf3a66f23.jpg |
0017e692d4f6f807_6d95 | Consider the real-world 3D locations of the objects. Which object has a lower location? | a woman wearing a blue jacket | a table with a vase of flowers on it | null | null | [
{
"bbox_3d": [
-0.3,
0.7,
1.7
],
"label": "a woman wearing a blue jacket"
},
{
"bbox_3d": [
0.3,
0.6,
1.5
],
"label": "a table with a vase of flowers on it"
}
] | [] | B | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a woman wearing a blue jacket is 1.3. The 3D height of a table with a vase of flowers on it is 0.7. The 3D height of a woman wearing a blue jacket is larger, meaning that the location of a woman wearing a blue jacket is higher. In other words, the location of a table with a vase of flowers on it is lower. Therefore, the answer is B. a woman wearing a blue jacket. | B. a woman wearing a blue jacket | height_higher | 0017e692d4f6f807.jpg |
0018315425ea8a20_17db | Consider the real-world 3D locations and orientations of the objects. If I stand at a yellow kayak with a red jacket's position facing where it is facing, is a man in a red jacket is standing on a boat on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
-0.3,
3.5,
12.4
],
"label": "a man in a red jacket is standing on a boat"
},
{
"bbox_3d": [
1.1,
0.4,
3.6
],
"label": "a yellow kayak with a red jacket"
}
] | [
{
"front_dir": [
-1,
0.2,
0.2
],
"label": "a yellow kayak with a red jacket",
"left_dir": [
0.2,
0.1,
1
]
}
] | A | To solve this problem, we first determine the 3D locations of a man in a red jacket is standing on a boat and a yellow kayak with a red jacket. Then we estimate the vector pointing from a yellow kayak with a red jacket to a man in a red jacket is standing on a boat, as well as the left direction of a yellow kayak with a red jacket. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man in a red jacket is standing on a boat is on the left of a yellow kayak with a red jacket. Otherwise, a man in a red jacket is standing on a boat is behind a yellow kayak with a red jacket. The 3D location of a man in a red jacket is standing on a boat is (-0.3, 3.5, 12.4). The 3D location of a yellow kayak with a red jacket is (1.1, 0.4, 3.6). The vector from a yellow kayak with a red jacket to a man in a red jacket is standing on a boat is hence (-1.4, 3.1, 8.8). The left direction of a yellow kayak with a red jacket is (0.2, 0.1, 1.0). The cosine similarity between the vector and the left direction is 0.90, corresponding to an angle of 25.26 degrees. The angle is smaller than 90 degrees, meaning that a man in a red jacket is standing on a boat is on the left of a yellow kayak with a red jacket. Therefore, the final answer is A. on the left. | A. on the left. | orientation_on_the_left | 0018315425ea8a20.jpg |
0018b3e86f60e214_00b1 | Consider the real-world 3D locations and orientations of the objects. If I stand at a wooden box with a mesh screen's position facing where it is facing, is a plant with green leaves on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
0.8,
0.1,
2.1
],
"label": "a plant with green leaves"
},
{
"bbox_3d": [
0,
0.3,
1.1
],
"label": "a wooden box with a mesh screen"
}
] | [
{
"front_dir": [
0.4,
0,
-0.9
],
"label": "a wooden box with a mesh screen",
"left_dir": [
-0.9,
0.1,
-0.4
]
}
] | B | To solve this problem, we first determine the 3D locations of a plant with green leaves and a wooden box with a mesh screen. Then we estimate the vector pointing from a wooden box with a mesh screen to a plant with green leaves, as well as the left direction of a wooden box with a mesh screen. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a plant with green leaves is on the left of a wooden box with a mesh screen. Otherwise, a plant with green leaves is behind a wooden box with a mesh screen. The 3D location of a plant with green leaves is (0.8, 0.1, 2.1). The 3D location of a wooden box with a mesh screen is (-0.0, 0.3, 1.1). The vector from a wooden box with a mesh screen to a plant with green leaves is hence (0.8, -0.2, 1.0). The left direction of a wooden box with a mesh screen is (-0.9, 0.1, -0.4). The cosine similarity between the vector and the left direction is -0.89, corresponding to an angle of 152.54 degrees. The angle is smaller than 90 degrees, meaning that a plant with green leaves is on the right of a wooden box with a mesh screen. Therefore, the final answer is B. on the right. | B. on the right. | orientation_on_the_left | 0018b3e86f60e214.jpg |
0018e237877764e9_eead | Consider the real-world 3D locations and orientations of the objects. Which side of a grey car on the road is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
0.9,
1.3,
22.5
],
"label": "a grey car on the road"
}
] | [
{
"front_dir": [
0,
0.1,
1
],
"label": "a grey car on the road",
"left_dir": [
1,
-0.1,
0
]
}
] | C | To solve this problem, we first estimate the 3D location of a grey car on the road. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a grey car on the road, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a grey car on the road that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a grey car on the road is (0.9, 1.3, 22.5). The vector from a grey car on the road to camera is hence (-0.9, -1.3, -22.5). The left direction of a grey car on the road is (1.0, -0.1, 0.0). The cosine similarity between the vector pointing to camera and the left direction is -0.06, corresponding to an angle of 93.66 degrees. Thus the angle between the vector pointing to camera and the right direction is 86.34 degrees. The front direction of a grey car on the road is (-0.0, 0.1, 1.0). The cosine similarity between the vector pointing to camera and the front direction is -1.00, corresponding to an angle of 175.93 degrees. Thus the angle between the vector pointing to camera and the back direction is 4.07 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 4.07 degrees. Thus the back side of a grey car on the road is facing the camera. Therefore, the final answer is C. back. | C. back. | orientation_viewpoint | 0018e237877764e9.jpg |
0018e9491515a4bb_e2b9 | Consider the real-world 3D locations and orientations of the objects. Which side of a white boat with a black motor is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-59.3,
-1.4,
264.7
],
"label": "a white boat with a black motor"
}
] | [
{
"front_dir": [
0.3,
-0.1,
-1
],
"label": "a white boat with a black motor",
"left_dir": [
-0.9,
0.1,
-0.3
]
}
] | A | To solve this problem, we first estimate the 3D location of a white boat with a black motor. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white boat with a black motor, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white boat with a black motor that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a white boat with a black motor is (-59.3, -1.4, 264.7). The vector from a white boat with a black motor to camera is hence (59.3, 1.4, -264.7). The left direction of a white boat with a black motor is (-0.9, 0.1, -0.3). The cosine similarity between the vector pointing to camera and the left direction is 0.08, corresponding to an angle of 85.65 degrees. Thus the angle between the vector pointing to camera and the right direction is 94.35 degrees. The front direction of a white boat with a black motor is (0.3, -0.1, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 1.00, corresponding to an angle of 5.49 degrees. Thus the angle between the vector pointing to camera and the back direction is 174.51 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.49 degrees. Thus the front side of a white boat with a black motor is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 0018e9491515a4bb.jpg |
001929c7ea49fa04_abb3 | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a cork with a brown color and a rough texture | a label with a brown color | null | null | [
{
"bbox_3d": [
-0.2,
0.4,
1
],
"label": "a cork with a brown color and a rough texture"
},
{
"bbox_3d": [
0.2,
0.1,
0.7
],
"label": "a label with a brown color"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a cork with a brown color and a rough texture and a label with a brown color. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a cork with a brown color and a rough texture is (-0.2, 0.4, 1.0). The 3D location of a label with a brown color is (0.2, 0.1, 0.7). The L2 distance from the camera to a cork with a brown color and a rough texture is 1.13. The L2 distance from the camera to a label with a brown color is 0.75. The distance to a cork with a brown color and a rough texture is larger. Therefore, the answer is A. a cork with a brown color and a rough texture. | A. a cork with a brown color and a rough texture. | location_closer_to_camera | 001929c7ea49fa04.jpg |
001950582b1a2690_ca5b | Consider the real-world 3D orientations of the objects. Are a black speaker and a stool with a white frame facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
-3.6,
0.9,
7.1
],
"label": "a black speaker"
},
{
"bbox_3d": [
-2.4,
0.3,
7.4
],
"label": "a stool with a white frame"
}
] | [
{
"front_dir": [
0.5,
-0.1,
-0.9
],
"label": "a black speaker",
"left_dir": [
-0.9,
0,
-0.5
]
},
{
"front_dir": [
0.4,
0,
-0.9
],
"label": "a stool with a white frame",
"left_dir": [
-0.9,
0,
-0.4
]
}
] | A | To solve this problem, we first detect the front directions of a black speaker and a stool with a white frame. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black speaker is (0.5, -0.1, -0.9). The front direction of a stool with a white frame is (0.4, 0.0, -0.9). The cosine similarity between the two front directions is 0.98, corresponding to an angle of 10.04. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions. | A. same or similar directions. | multi_object_same_direction | 001950582b1a2690.jpg |
001952f2e3bf13a5_9a6d | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a white typewriter with a black keyboard and a white keyboard with black keys, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.3,
0.7,
1.1
],
"label": "a white typewriter with a black keyboard"
},
{
"bbox_3d": [
-0.2,
0.5,
0.9
],
"label": "a white keyboard with black keys"
}
] | [
{
"front_dir": [
0,
-0.9,
-0.4
],
"label": "a white typewriter with a black keyboard",
"left_dir": [
-1,
0.1,
-0.3
]
},
{
"front_dir": [
0.3,
-0.9,
-0.3
],
"label": "a white keyboard with black keys",
"left_dir": [
-0.9,
-0.1,
-0.5
]
}
] | A | To solve this problem, we first detect the front directions of a white typewriter with a black keyboard and a white keyboard with black keys. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a white typewriter with a black keyboard is (0.0, -0.9, -0.4). The front direction of a white keyboard with black keys is (0.3, -0.9, -0.3). The cosine similarity between the two front directions is 0.95, corresponding to an angle of 18.15. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 001952f2e3bf13a5.jpg |
001952f2e3bf13a5_beaa | Consider the real-world 3D locations and orientations of the objects. Which object is a white desktop computer facing towards, a white keyboard with black keys or the a white computer with a black screen? | a white keyboard with black keys | a white computer with a black screen | null | null | [
{
"bbox_3d": [
-0.4,
0.7,
0.9
],
"label": "a white desktop computer"
},
{
"bbox_3d": [
-0.2,
0.5,
0.9
],
"label": "a white keyboard with black keys"
},
{
"bbox_3d": [
-0.3,
0.7,
0.8
],
"label": "a white computer with a black screen"
}
] | [
{
"front_dir": [
0.4,
-0.9,
-0.2
],
"label": "a white desktop computer",
"left_dir": [
-0.8,
-0.2,
-0.6
]
},
{
"front_dir": [
0.3,
-0.9,
-0.3
],
"label": "a white keyboard with black keys",
"left_dir": [
-0.9,
-0.1,
-0.5
]
},
{
"front_dir": [
0.4,
-0.9,
-0.2
],
"label": "a white computer with a black screen",
"left_dir": [
-0.8,
-0.3,
-0.5
]
}
] | A | To solve this problem, we first detect the 3D location of a white desktop computer, a white keyboard with black keys, and a white computer with a black screen. Then we compute the cosine similarities between the front direction of a white desktop computer and the vectors from a white desktop computer to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a white desktop computer is facing towards. The 3D location of a white desktop computer is (-0.4, 0.7, 0.9). The 3D location of a white keyboard with black keys is (-0.2, 0.5, 0.9). The 3D location of a white computer with a black screen is (-0.3, 0.7, 0.8). The front direction of a white desktop computer is (0.4, -0.9, -0.2). First we consider if a white desktop computer is facing towards the a white keyboard with black keys. The vector from a white desktop computer to a white keyboard with black keys is (0.2, -0.2, -0.0). The cosine similarity between the front direction and the vector is 0.97, corresponding to an angle of 14.46 degrees. First we consider if a white desktop computer is facing towards the a white computer with a black screen. The vector from a white desktop computer to a white computer with a black screen is (0.1, 0.0, -0.1). The cosine similarity between the front direction and the vector is 0.44, corresponding to an angle of 63.61 degrees. We find that the angle between the front direction and a white keyboard with black keys is smaller. Therefore, the final answer is A. a white keyboard with black keys. | A. a white keyboard with black keys. | multi_object_facing | 001952f2e3bf13a5.jpg |
0019699a08c6bdff_1b42 | Consider the real-world 3D locations of the objects. Are the a black and white photo of two men and the a man in a black and white striped suit next to each other or far away from each other? | next to each other | far away from each other | null | null | [
{
"bbox_3d": [
0.2,
0.5,
1.1
],
"label": "a black and white photo of two men"
},
{
"bbox_3d": [
-0.2,
0.5,
1.1
],
"label": "a man in a black and white striped suit"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a black and white photo of two men and a man in a black and white striped suit. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a black and white photo of two men is (0.2, 0.5, 1.1). The 3D location of a man in a black and white striped suit is (-0.2, 0.5, 1.1). The L2 distance between the two objects is 0.42. The size of the a black and white photo of two men is roughly 0.85. The size of the a man in a black and white striped suit is roughly 0.75. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other. | A. next to each other. | location_next_to | 0019699a08c6bdff.jpg |
001a09c6d3585b52_510d | Consider the real-world 3D locations of the objects. Is a man wearing a backpack and a helmet directly above a rope? | yes | no | null | null | [
{
"bbox_3d": [
-0.9,
17.5,
3.5
],
"label": "a man wearing a backpack and a helmet"
},
{
"bbox_3d": [
-1.2,
16.7,
3.3
],
"label": "a rope"
}
] | [] | A | To solve this problem, we first determine the 3D locations of a man wearing a backpack and a helmet and a rope. Then we compute the vector pointing from a rope to a man wearing a backpack and a helmet, as well as the up direction of a rope. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a man wearing a backpack and a helmet is directly above a rope. Otherwise, then a man wearing a backpack and a helmet is not directly above a rope. The 3D location of a man wearing a backpack and a helmet is (-0.9, 17.5, 3.5). The 3D location of a rope is (-1.2, 16.7, 3.3). The vector from a rope to a man wearing a backpack and a helmet is hence (0.3, 0.8, 0.2). The up direction of a rope is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.91, corresponding to an angle of 24 degrees. The angle between the vector and the up direction is small, meaning that a man wearing a backpack and a helmet is directly above a rope. Therefore, the answer is A. yes. | A. yes. | location_above | 001a09c6d3585b52.jpg |
001a0d77808c53d0_1fa6 | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a girl with a black leather jacket | a black stool | null | null | [
{
"bbox_3d": [
0.4,
1.3,
1
],
"label": "a girl with a black leather jacket"
},
{
"bbox_3d": [
0.3,
0.8,
0.9
],
"label": "a black stool"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a girl with a black leather jacket and a black stool. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a girl with a black leather jacket is (0.4, 1.3, 1.0). The 3D location of a black stool is (0.3, 0.8, 0.9). The L2 distance from the camera to a girl with a black leather jacket is 1.71. The L2 distance from the camera to a black stool is 1.27. The distance to a girl with a black leather jacket is larger. Therefore, the answer is A. a girl with a black leather jacket. | A. a girl with a black leather jacket. | location_closer_to_camera | 001a0d77808c53d0.jpg |
001c7ecaa2019715_f80c | Consider the real-world 3D locations and orientations of the objects. Which object is a staircase with a red railing facing towards, a wooden balustrade or the a green and white toy? | a wooden balustrade | a green and white toy | null | null | [
{
"bbox_3d": [
0.2,
0.8,
9.3
],
"label": "a staircase with a red railing"
},
{
"bbox_3d": [
5.7,
1.8,
12.1
],
"label": "a wooden balustrade"
},
{
"bbox_3d": [
0,
0.8,
1.8
],
"label": "a green and white toy"
}
] | [
{
"front_dir": [
0.1,
0.1,
-1
],
"label": "a staircase with a red railing",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | B | To solve this problem, we first detect the 3D location of a staircase with a red railing, a wooden balustrade, and a green and white toy. Then we compute the cosine similarities between the front direction of a staircase with a red railing and the vectors from a staircase with a red railing to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a staircase with a red railing is facing towards. The 3D location of a staircase with a red railing is (0.2, 0.8, 9.3). The 3D location of a wooden balustrade is (5.7, 1.8, 12.1). The 3D location of a green and white toy is (0.0, 0.8, 1.8). The front direction of a staircase with a red railing is (0.1, 0.1, -1.0). First we consider if a staircase with a red railing is facing towards the a wooden balustrade. The vector from a staircase with a red railing to a wooden balustrade is (5.5, 1.0, 2.8). The cosine similarity between the front direction and the vector is -0.36, corresponding to an angle of 111.01 degrees. First we consider if a staircase with a red railing is facing towards the a green and white toy. The vector from a staircase with a red railing to a green and white toy is (-0.2, -0.0, -7.5). The cosine similarity between the front direction and the vector is 0.98, corresponding to an angle of 10.23 degrees. We find that the angle between the front direction and a green and white toy is smaller. Therefore, the final answer is B. a green and white toy. | B. a green and white toy. | multi_object_facing | 001c7ecaa2019715.jpg |
001d36e16e301fe2_0942 | Consider the real-world 3D orientations of the objects. Are a small boat with a flag on it and a concrete dock facing same or similar directions, or very different directions? | same or similar directions | very different directions | null | null | [
{
"bbox_3d": [
6.3,
1.7,
66.2
],
"label": "a small boat with a flag on it"
},
{
"bbox_3d": [
-18.7,
2.9,
112.6
],
"label": "a concrete dock"
}
] | [
{
"front_dir": [
-1,
0.1,
0
],
"label": "a small boat with a flag on it",
"left_dir": [
0,
0.1,
1
]
},
{
"front_dir": [
0.2,
-0.1,
-1
],
"label": "a concrete dock",
"left_dir": [
-1,
0.1,
-0.2
]
}
] | B | To solve this problem, we first detect the front directions of a small boat with a flag on it and a concrete dock. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a small boat with a flag on it is (-1.0, 0.1, 0.0). The front direction of a concrete dock is (0.2, -0.1, -1.0). The cosine similarity between the two front directions is -0.26, corresponding to an angle of 104.88. The angle is large, meaning that the two objects are facing very different directions. Therefore, the final answer is B. very different directions. | B. very different directions. | multi_object_same_direction | 001d36e16e301fe2.jpg |
001d36e16e301fe2_543a | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a small boat with a flag on it and a large blue barge, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
6.3,
1.7,
66.2
],
"label": "a small boat with a flag on it"
},
{
"bbox_3d": [
5.4,
6.9,
127
],
"label": "a large blue barge"
}
] | [
{
"front_dir": [
-1,
0.1,
0
],
"label": "a small boat with a flag on it",
"left_dir": [
0,
0.1,
1
]
},
{
"front_dir": [
0,
-0.1,
-1
],
"label": "a large blue barge",
"left_dir": [
-1,
-0.1,
0
]
}
] | B | To solve this problem, we first detect the front directions of a small boat with a flag on it and a large blue barge. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a small boat with a flag on it is (-1.0, 0.1, 0.0). The front direction of a large blue barge is (0.0, -0.1, -1.0). The cosine similarity between the two front directions is -0.07, corresponding to an angle of 93.76. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular. | B. perpendicular. | multi_object_parallel | 001d36e16e301fe2.jpg |
001d36e16e301fe2_7c26 | Consider the real-world 3D locations and orientations of the objects. Which object is a large blue barge facing towards, a concrete dock or the a small boat with a flag on it? | a concrete dock | a small boat with a flag on it | null | null | [
{
"bbox_3d": [
5.4,
6.9,
127
],
"label": "a large blue barge"
},
{
"bbox_3d": [
-18.7,
2.9,
112.6
],
"label": "a concrete dock"
},
{
"bbox_3d": [
6.3,
1.7,
66.2
],
"label": "a small boat with a flag on it"
}
] | [
{
"front_dir": [
0,
-0.1,
-1
],
"label": "a large blue barge",
"left_dir": [
-1,
-0.1,
0
]
},
{
"front_dir": [
0.2,
-0.1,
-1
],
"label": "a concrete dock",
"left_dir": [
-1,
0.1,
-0.2
]
},
{
"front_dir": [
-1,
0.1,
0
],
"label": "a small boat with a flag on it",
"left_dir": [
0,
0.1,
1
]
}
] | B | To solve this problem, we first detect the 3D location of a large blue barge, a concrete dock, and a small boat with a flag on it. Then we compute the cosine similarities between the front direction of a large blue barge and the vectors from a large blue barge to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a large blue barge is facing towards. The 3D location of a large blue barge is (5.4, 6.9, 127.0). The 3D location of a concrete dock is (-18.7, 2.9, 112.6). The 3D location of a small boat with a flag on it is (6.3, 1.7, 66.2). The front direction of a large blue barge is (0.0, -0.1, -1.0). First we consider if a large blue barge is facing towards the a concrete dock. The vector from a large blue barge to a concrete dock is (-24.1, -4.0, -14.5). The cosine similarity between the front direction and the vector is 0.49, corresponding to an angle of 60.76 degrees. First we consider if a large blue barge is facing towards the a small boat with a flag on it. The vector from a large blue barge to a small boat with a flag on it is (0.9, -5.2, -60.8). The cosine similarity between the front direction and the vector is 1.00, corresponding to an angle of 2.50 degrees. We find that the angle between the front direction and a small boat with a flag on it is smaller. Therefore, the final answer is B. a small boat with a flag on it. | B. a small boat with a flag on it. | multi_object_facing | 001d36e16e301fe2.jpg |
001e040e9f8a2d4f_3c1e | Consider the real-world 3D locations and orientations of the objects. If I stand at a boat made of wood's position facing where it is facing, is a pink flower with green leaves in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
-3.1,
3.6,
13.3
],
"label": "a pink flower with green leaves"
},
{
"bbox_3d": [
-0.1,
2.4,
5.3
],
"label": "a boat made of wood"
}
] | [
{
"front_dir": [
0.1,
-0.1,
-1
],
"label": "a boat made of wood",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | B | To solve this problem, we first determine the 3D locations of a pink flower with green leaves and a boat made of wood. Then we estimate the vector pointing from a boat made of wood to a pink flower with green leaves, as well as the front direction of a boat made of wood. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a pink flower with green leaves is in front of a boat made of wood. Otherwise, a pink flower with green leaves is behind a boat made of wood. The 3D location of a pink flower with green leaves is (-3.1, 3.6, 13.3). The 3D location of a boat made of wood is (-0.1, 2.4, 5.3). The vector from a boat made of wood to a pink flower with green leaves is hence (-3.0, 1.2, 7.9). The front direction of a boat made of wood is (0.1, -0.1, -1.0). The cosine similarity between the vector and the front direction is -0.96, corresponding to an angle of 164.17 degrees. The angle is smaller than 90 degrees, meaning that a pink flower with green leaves is behind a boat made of wood. Therefore, the final answer is B. behind. | B. behind. | orientation_in_front_of | 001e040e9f8a2d4f.jpg |
001e040e9f8a2d4f_81e3 | Consider the real-world 3D locations and orientations of the objects. Which side of a boat made of wood is facing a pink flower with green leaves? | front | left | back | right | [
{
"bbox_3d": [
-0.1,
2.4,
5.3
],
"label": "a boat made of wood"
},
{
"bbox_3d": [
-3.1,
3.6,
13.3
],
"label": "a pink flower with green leaves"
}
] | [
{
"front_dir": [
0.1,
-0.1,
-1
],
"label": "a boat made of wood",
"left_dir": [
-1,
0.1,
-0.1
]
}
] | C | To solve this problem, we first detect the 3D locations of a boat made of wood and a pink flower with green leaves. Then we compute the vector pointing from a boat made of wood to a pink flower with green leaves. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a boat made of wood, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a boat made of wood that is facing a pink flower with green leaves corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a boat made of wood is (-0.1, 2.4, 5.3). The 3D location of a pink flower with green leaves is (-3.1, 3.6, 13.3). The vector from a boat made of wood to a pink flower with green leaves is hence (-3.0, 1.2, 7.9). The left direction of a boat made of wood is (-1.0, 0.1, -0.1). The cosine similarity between the vector pointing to a pink flower with green leaves and the left direction is 0.27, corresponding to an angle of 74.23 degrees. Thus the angle between the vector pointing to a pink flower with green leaves and the right direction is 105.77 degrees. The front direction of a boat made of wood is (0.1, -0.1, -1.0). The cosine similarity between the vector pointing to a pink flower with green leaves and the front direction is -0.96, corresponding to an angle of 164.17 degrees. Thus the angle between the vector pointing to a pink flower with green leaves and the back direction is 15.83 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 15.83 degrees. Thus the back side of a boat made of wood is facing the a pink flower with green leaves. Therefore, the final answer is C. back. | C. back. | multi_object_viewpoint_towards_object | 001e040e9f8a2d4f.jpg |
001e6e78fbbbcdb4_b14e | Consider the real-world 3D locations and orientations of the objects. Which side of a small car on display is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-0.3,
4.1,
7
],
"label": "a small car on display"
}
] | [
{
"front_dir": [
0.4,
-0.5,
-0.7
],
"label": "a small car on display",
"left_dir": [
-0.8,
0.1,
-0.5
]
}
] | A | To solve this problem, we first estimate the 3D location of a small car on display. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a small car on display, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a small car on display that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a small car on display is (-0.3, 4.1, 7.0). The vector from a small car on display to camera is hence (0.3, -4.1, -7.0). The left direction of a small car on display is (-0.8, 0.1, -0.5). The cosine similarity between the vector pointing to camera and the left direction is 0.39, corresponding to an angle of 66.88 degrees. Thus the angle between the vector pointing to camera and the right direction is 113.12 degrees. The front direction of a small car on display is (0.4, -0.5, -0.7). The cosine similarity between the vector pointing to camera and the front direction is 0.92, corresponding to an angle of 23.40 degrees. Thus the angle between the vector pointing to camera and the back direction is 156.60 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 23.40 degrees. Thus the front side of a small car on display is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 001e6e78fbbbcdb4.jpg |
001eb9474d2cd1f2_4f68 | Consider the real-world 3D locations of the objects. Which is closer to a cart with a picture on it, a white jumpsuit or a person wearing a white suit? | a white jumpsuit | a person wearing a white suit | null | null | [
{
"bbox_3d": [
0.8,
0.2,
2.8
],
"label": "a cart with a picture on it"
},
{
"bbox_3d": [
-0.4,
1.9,
3.1
],
"label": "a white jumpsuit"
},
{
"bbox_3d": [
-0.5,
1.1,
3.1
],
"label": "a person wearing a white suit"
}
] | [] | B | To solve this problem, we first detect the 3D location of a cart with a picture on it, a white jumpsuit, and a person wearing a white suit. Then we compute the L2 distances between a cart with a picture on it and a white jumpsuit, and between a cart with a picture on it and a person wearing a white suit. The object that is closer to a cart with a picture on it is the one with a smaller distance. The 3D location of a cart with a picture on it is (0.8, 0.2, 2.8). The 3D location of a white jumpsuit is (-0.4, 1.9, 3.1). The 3D location of a person wearing a white suit is (-0.5, 1.1, 3.1). The L2 distance between a cart with a picture on it and a white jumpsuit is 2.136788725387188. The L2 distance between a cart with a picture on it and a person wearing a white suit is 1.6144927332784667. Between the two distances, the distance between a cart with a picture on it and a person wearing a white suit is smaller. Therefore, the final answer is B. a person wearing a white suit. | B. a person wearing a white suit. | multi_object_closer_to | 001eb9474d2cd1f2.jpg |
001f2bec02f16977_5723 | Consider the real-world 3D locations and orientations of the objects. If I stand at a wooden bulletin board with a picture of a cat's position facing where it is facing, is a store that sells paintings on the left or right of me? | on the left | on the right | null | null | [
{
"bbox_3d": [
-0.8,
0.6,
1.7
],
"label": "a store that sells paintings"
},
{
"bbox_3d": [
0.2,
0.5,
1.2
],
"label": "a wooden bulletin board with a picture of a cat"
}
] | [
{
"front_dir": [
-0.1,
-0.1,
-1
],
"label": "a wooden bulletin board with a picture of a cat",
"left_dir": [
-1,
0.1,
0.1
]
}
] | A | To solve this problem, we first determine the 3D locations of a store that sells paintings and a wooden bulletin board with a picture of a cat. Then we estimate the vector pointing from a wooden bulletin board with a picture of a cat to a store that sells paintings, as well as the left direction of a wooden bulletin board with a picture of a cat. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a store that sells paintings is on the left of a wooden bulletin board with a picture of a cat. Otherwise, a store that sells paintings is behind a wooden bulletin board with a picture of a cat. The 3D location of a store that sells paintings is (-0.8, 0.6, 1.7). The 3D location of a wooden bulletin board with a picture of a cat is (0.2, 0.5, 1.2). The vector from a wooden bulletin board with a picture of a cat to a store that sells paintings is hence (-1.0, 0.1, 0.5). The left direction of a wooden bulletin board with a picture of a cat is (-1.0, 0.1, 0.1). The cosine similarity between the vector and the left direction is 0.95, corresponding to an angle of 18.78 degrees. The angle is smaller than 90 degrees, meaning that a store that sells paintings is on the left of a wooden bulletin board with a picture of a cat. Therefore, the final answer is A. on the left. | A. on the left. | orientation_on_the_left | 001f2bec02f16977.jpg |
001f2bec02f16977_2d2f | Consider the real-world 3D locations and orientations of the objects. Which side of a wooden bulletin board with a picture of a cat is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
0.2,
0.5,
1.2
],
"label": "a wooden bulletin board with a picture of a cat"
}
] | [
{
"front_dir": [
-0.1,
-0.1,
-1
],
"label": "a wooden bulletin board with a picture of a cat",
"left_dir": [
-1,
0.1,
0.1
]
}
] | A | To solve this problem, we first estimate the 3D location of a wooden bulletin board with a picture of a cat. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden bulletin board with a picture of a cat, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden bulletin board with a picture of a cat that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a wooden bulletin board with a picture of a cat is (0.2, 0.5, 1.2). The vector from a wooden bulletin board with a picture of a cat to camera is hence (-0.2, -0.5, -1.2). The left direction of a wooden bulletin board with a picture of a cat is (-1.0, 0.1, 0.1). The cosine similarity between the vector pointing to camera and the left direction is 0.04, corresponding to an angle of 87.50 degrees. Thus the angle between the vector pointing to camera and the right direction is 92.50 degrees. The front direction of a wooden bulletin board with a picture of a cat is (-0.1, -0.1, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.95, corresponding to an angle of 17.54 degrees. Thus the angle between the vector pointing to camera and the back direction is 162.46 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 17.54 degrees. Thus the front side of a wooden bulletin board with a picture of a cat is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 001f2bec02f16977.jpg |
001faa232f9110f8_a03d | Consider the real-world 3D locations of the objects. Is a man in a plaid shirt directly above a hand with a brown skin tone? | yes | no | null | null | [
{
"bbox_3d": [
0,
3.3,
2.1
],
"label": "a man in a plaid shirt"
},
{
"bbox_3d": [
0.2,
3,
2.2
],
"label": "a hand with a brown skin tone"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a man in a plaid shirt and a hand with a brown skin tone. Then we compute the vector pointing from a hand with a brown skin tone to a man in a plaid shirt, as well as the up direction of a hand with a brown skin tone. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a man in a plaid shirt is directly above a hand with a brown skin tone. Otherwise, then a man in a plaid shirt is not directly above a hand with a brown skin tone. The 3D location of a man in a plaid shirt is (0.0, 3.3, 2.1). The 3D location of a hand with a brown skin tone is (0.2, 3.0, 2.2). The vector from a hand with a brown skin tone to a man in a plaid shirt is hence (-0.2, 0.2, -0.1). The up direction of a hand with a brown skin tone is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.73, corresponding to an angle of 43 degrees. The angle between the vector and the up direction is large, meaning that a man in a plaid shirt is not directly above a hand with a brown skin tone. Therefore, the answer is B. no. | B. no. | location_above | 001faa232f9110f8.jpg |
001fc590a2888f48_02b2 | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a man in a yellow and white shirt riding a mountain bike and a dirt bike on a dirt road, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.3,
1.3,
6.1
],
"label": "a man in a yellow and white shirt riding a mountain bike"
},
{
"bbox_3d": [
0.3,
1.3,
2.8
],
"label": "a dirt bike on a dirt road"
}
] | [
{
"front_dir": [
-0.2,
-0.1,
-1
],
"label": "a man in a yellow and white shirt riding a mountain bike",
"left_dir": [
-1,
0.2,
0.2
]
},
{
"front_dir": [
0,
0.3,
-1
],
"label": "a dirt bike on a dirt road",
"left_dir": [
-1,
0.2,
0.1
]
}
] | A | To solve this problem, we first detect the front directions of a man in a yellow and white shirt riding a mountain bike and a dirt bike on a dirt road. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a man in a yellow and white shirt riding a mountain bike is (-0.2, -0.1, -1.0). The front direction of a dirt bike on a dirt road is (-0.0, 0.3, -1.0). The cosine similarity between the two front directions is 0.94, corresponding to an angle of 19.29. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 001fc590a2888f48.jpg |
00200ddc8b80344f_c38d | Consider the real-world 3D locations of the objects. Is a yellow bead with a blue design directly underneath a yellow and green beaded necklace? | yes | no | null | null | [
{
"bbox_3d": [
0,
0.4,
0.5
],
"label": "a yellow and green beaded necklace"
},
{
"bbox_3d": [
0,
0.3,
0.4
],
"label": "a yellow bead with a blue design"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a yellow and green beaded necklace and a yellow bead with a blue design. Then we compute the vector pointing from a yellow bead with a blue design to a yellow and green beaded necklace, as well as the up direction of a yellow bead with a blue design. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a yellow and green beaded necklace is directly above a yellow bead with a blue design. Otherwise, then a yellow and green beaded necklace is not directly above a yellow bead with a blue design. To solve the question, we first determine if a yellow and green beaded necklace is directly above a yellow bead with a blue design. The 3D location of a yellow and green beaded necklace is (0.0, 0.4, 0.5). The 3D location of a yellow bead with a blue design is (0.0, 0.3, 0.4). The vector from a yellow bead with a blue design to a yellow and green beaded necklace is hence (0.0, 0.1, 0.1). The up direction of a yellow bead with a blue design is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.77, corresponding to an angle of 39 degrees. The angle between the vector and the up direction is large, meaning that a yellow and green beaded necklace is not directly above a yellow bead with a blue design. In other words, a yellow bead with a blue design is not directly underneath a yellow and green beaded necklace. Therefore, the answer is B. no. | B. no. | location_above | 00200ddc8b80344f.jpg |
0020a55549ab155b_224a | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a dock with a boat and a large orange and white boat, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
0.9,
11,
216.5
],
"label": "a dock with a boat"
},
{
"bbox_3d": [
5.6,
3.8,
107.4
],
"label": "a large orange and white boat"
}
] | [
{
"front_dir": [
0.1,
-0.1,
-1
],
"label": "a dock with a boat",
"left_dir": [
-1,
0,
-0.1
]
},
{
"front_dir": [
0,
0.1,
-1
],
"label": "a large orange and white boat",
"left_dir": [
-1,
0,
0
]
}
] | A | To solve this problem, we first detect the front directions of a dock with a boat and a large orange and white boat. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a dock with a boat is (0.1, -0.1, -1.0). The front direction of a large orange and white boat is (0.0, 0.1, -1.0). The cosine similarity between the two front directions is 0.98, corresponding to an angle of 11.14. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel. | A. parallel. | multi_object_parallel | 0020a55549ab155b.jpg |
0020affb95ae70b8_5eb1 | Consider the real-world 3D location of the objects. Which object is further away from the camera? | a large oak tree with a twisted trunk | a road with a curb | null | null | [
{
"bbox_3d": [
1.8,
4.9,
6.2
],
"label": "a large oak tree with a twisted trunk"
},
{
"bbox_3d": [
0,
0.1,
5.2
],
"label": "a road with a curb"
}
] | [] | A | To solve this problem, we first estimate the 3D locations of a large oak tree with a twisted trunk and a road with a curb. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a large oak tree with a twisted trunk is (1.8, 4.9, 6.2). The 3D location of a road with a curb is (-0.0, 0.1, 5.2). The L2 distance from the camera to a large oak tree with a twisted trunk is 8.11. The L2 distance from the camera to a road with a curb is 5.22. The distance to a large oak tree with a twisted trunk is larger. Therefore, the answer is A. a large oak tree with a twisted trunk. | A. a large oak tree with a twisted trunk. | location_closer_to_camera | 0020affb95ae70b8.jpg |
0020ca4d32839638_0239 | Consider the real-world 3D locations of the objects. Which object has a higher location? | a group of people sitting in a booth | a glass of water | null | null | [
{
"bbox_3d": [
0.3,
1.1,
2.4
],
"label": "a group of people sitting in a booth"
},
{
"bbox_3d": [
0.6,
0.2,
1.3
],
"label": "a glass of water"
}
] | [] | A | To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a group of people sitting in a booth is 3.1. The 3D height of a glass of water is 0.4. The 3D height of a group of people sitting in a booth is larger, meaning that the location of a group of people sitting in a booth is higher. Therefore, the answer is A. a glass of water. | A. a glass of water. | height_higher | 0020ca4d32839638.jpg |
0020f89e4cb5041d_df52 | Consider the real-world 3D locations of the objects. Is a white cup filled with fries directly above a tomato? | yes | no | null | null | [
{
"bbox_3d": [
-0.1,
0.2,
0.3
],
"label": "a white cup filled with fries"
},
{
"bbox_3d": [
0,
0.1,
0.2
],
"label": "a tomato"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a white cup filled with fries and a tomato. Then we compute the vector pointing from a tomato to a white cup filled with fries, as well as the up direction of a tomato. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a white cup filled with fries is directly above a tomato. Otherwise, then a white cup filled with fries is not directly above a tomato. The 3D location of a white cup filled with fries is (-0.1, 0.2, 0.3). The 3D location of a tomato is (0.0, 0.1, 0.2). The vector from a tomato to a white cup filled with fries is hence (-0.2, 0.1, 0.1). The up direction of a tomato is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.33, corresponding to an angle of 70 degrees. The angle between the vector and the up direction is large, meaning that a white cup filled with fries is not directly above a tomato. Therefore, the answer is B. no. | B. no. | location_above | 0020f89e4cb5041d.jpg |
002136e5f6aefe6b_ea4e | Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a car with a hose connected to the tire and a white race car, parallel of perpendicular to each other? | parallel | perpendicular | null | null | [
{
"bbox_3d": [
-0.6,
1,
2.2
],
"label": "a car with a hose connected to the tire"
},
{
"bbox_3d": [
0.4,
0.9,
2.1
],
"label": "a white race car"
}
] | [
{
"front_dir": [
0.9,
0,
0.3
],
"label": "a car with a hose connected to the tire",
"left_dir": [
0.3,
-0.2,
-0.9
]
},
{
"front_dir": [
0.4,
-0.2,
-0.9
],
"label": "a white race car",
"left_dir": [
-0.9,
-0.1,
-0.4
]
}
] | B | To solve this problem, we first detect the front directions of a car with a hose connected to the tire and a white race car. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a car with a hose connected to the tire is (0.9, 0.0, 0.3). The front direction of a white race car is (0.4, -0.2, -0.9). The cosine similarity between the two front directions is 0.03, corresponding to an angle of 88.41. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular. | B. perpendicular. | multi_object_parallel | 002136e5f6aefe6b.jpg |
00219cfe531873f2_56a6 | Consider the real-world 3D locations of the objects. Is a man wearing a red shirt directly underneath a door with a metal grill? | yes | no | null | null | [
{
"bbox_3d": [
0.5,
1.7,
2.2
],
"label": "a door with a metal grill"
},
{
"bbox_3d": [
0.1,
1.1,
3
],
"label": "a man wearing a red shirt"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a door with a metal grill and a man wearing a red shirt. Then we compute the vector pointing from a man wearing a red shirt to a door with a metal grill, as well as the up direction of a man wearing a red shirt. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a door with a metal grill is directly above a man wearing a red shirt. Otherwise, then a door with a metal grill is not directly above a man wearing a red shirt. To solve the question, we first determine if a door with a metal grill is directly above a man wearing a red shirt. The 3D location of a door with a metal grill is (0.5, 1.7, 2.2). The 3D location of a man wearing a red shirt is (0.1, 1.1, 3.0). The vector from a man wearing a red shirt to a door with a metal grill is hence (0.4, 0.6, -0.9). The up direction of a man wearing a red shirt is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.55, corresponding to an angle of 56 degrees. The angle between the vector and the up direction is large, meaning that a door with a metal grill is not directly above a man wearing a red shirt. In other words, a man wearing a red shirt is not directly underneath a door with a metal grill. Therefore, the answer is B. no. | B. no. | location_above | 00219cfe531873f2.jpg |
0021b181c2993928_a98b | Consider the real-world 3D locations and orientations of the objects. Which side of a black baby carriage is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-0.1,
1.1,
5.2
],
"label": "a black baby carriage"
}
] | [
{
"front_dir": [
-0.2,
0,
-1
],
"label": "a black baby carriage",
"left_dir": [
-1,
0.2,
0.2
]
}
] | A | To solve this problem, we first estimate the 3D location of a black baby carriage. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black baby carriage, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black baby carriage that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a black baby carriage is (-0.1, 1.1, 5.2). The vector from a black baby carriage to camera is hence (0.1, -1.1, -5.2). The left direction of a black baby carriage is (-1.0, 0.2, 0.2). The cosine similarity between the vector pointing to camera and the left direction is -0.29, corresponding to an angle of 107.03 degrees. Thus the angle between the vector pointing to camera and the right direction is 72.97 degrees. The front direction of a black baby carriage is (-0.2, 0.0, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.94, corresponding to an angle of 19.30 degrees. Thus the angle between the vector pointing to camera and the back direction is 160.70 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 19.30 degrees. Thus the front side of a black baby carriage is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 0021b181c2993928.jpg |
0021d61e016ba1bc_ac0e | Consider the real-world 3D locations and orientations of the objects. Which side of a black stool is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-0.3,
0.8,
4.7
],
"label": "a black stool"
}
] | [
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a black stool",
"left_dir": [
-0.9,
0.1,
-0.4
]
}
] | A | To solve this problem, we first estimate the 3D location of a black stool. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black stool, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black stool that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a black stool is (-0.3, 0.8, 4.7). The vector from a black stool to camera is hence (0.3, -0.8, -4.7). The left direction of a black stool is (-0.9, 0.1, -0.4). The cosine similarity between the vector pointing to camera and the left direction is 0.35, corresponding to an angle of 69.58 degrees. Thus the angle between the vector pointing to camera and the right direction is 110.42 degrees. The front direction of a black stool is (0.4, -0.1, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 0.93, corresponding to an angle of 21.36 degrees. Thus the angle between the vector pointing to camera and the back direction is 158.64 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 21.36 degrees. Thus the front side of a black stool is facing the camera. Therefore, the final answer is A. front. | A. front. | orientation_viewpoint | 0021d61e016ba1bc.jpg |
0022148abedec60d_a35c | Consider the real-world 3D locations and orientations of the objects. If I stand at a black and white photo of a car's position facing where it is facing, is a man in white pants in front of me or behind me? | in front of | behind | null | null | [
{
"bbox_3d": [
-0.3,
1,
1.4
],
"label": "a man in white pants"
},
{
"bbox_3d": [
0,
1,
1.4
],
"label": "a black and white photo of a car"
}
] | [
{
"front_dir": [
-0.9,
0,
-0.5
],
"label": "a black and white photo of a car",
"left_dir": [
-0.5,
0.3,
0.8
]
}
] | A | To solve this problem, we first determine the 3D locations of a man in white pants and a black and white photo of a car. Then we estimate the vector pointing from a black and white photo of a car to a man in white pants, as well as the front direction of a black and white photo of a car. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man in white pants is in front of a black and white photo of a car. Otherwise, a man in white pants is behind a black and white photo of a car. The 3D location of a man in white pants is (-0.3, 1.0, 1.4). The 3D location of a black and white photo of a car is (0.0, 1.0, 1.4). The vector from a black and white photo of a car to a man in white pants is hence (-0.3, 0.0, 0.0). The front direction of a black and white photo of a car is (-0.9, -0.0, -0.5). The cosine similarity between the vector and the front direction is 0.84, corresponding to an angle of 33.32 degrees. The angle is smaller than 90 degrees, meaning that a man in white pants is in front of a black and white photo of a car. Therefore, the final answer is A. in front of. | A. in front of. | orientation_in_front_of | 0022148abedec60d.jpg |
002223b533d730f5_6772 | Consider the real-world 3D locations and orientations of the objects. Which object is a white van facing towards, a road with cars parked on the side or the a snow covered mountain? | a road with cars parked on the side | a snow covered mountain | null | null | [
{
"bbox_3d": [
-6.3,
1.5,
33
],
"label": "a white van"
},
{
"bbox_3d": [
-3.4,
0.5,
23.4
],
"label": "a road with cars parked on the side"
},
{
"bbox_3d": [
-4,
35.4,
136.1
],
"label": "a snow covered mountain"
}
] | [
{
"front_dir": [
0.7,
-0.1,
-0.7
],
"label": "a white van",
"left_dir": [
-0.7,
0,
-0.7
]
}
] | A | To solve this problem, we first detect the 3D location of a white van, a road with cars parked on the side, and a snow covered mountain. Then we compute the cosine similarities between the front direction of a white van and the vectors from a white van to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a white van is facing towards. The 3D location of a white van is (-6.3, 1.5, 33.0). The 3D location of a road with cars parked on the side is (-3.4, 0.5, 23.4). The 3D location of a snow covered mountain is (-4.0, 35.4, 136.1). The front direction of a white van is (0.7, -0.1, -0.7). First we consider if a white van is facing towards the a road with cars parked on the side. The vector from a white van to a road with cars parked on the side is (2.9, -1.1, -9.6). The cosine similarity between the front direction and the vector is 0.90, corresponding to an angle of 26.20 degrees. First we consider if a white van is facing towards the a snow covered mountain. The vector from a white van to a snow covered mountain is (2.3, 33.8, 103.2). The cosine similarity between the front direction and the vector is -0.71, corresponding to an angle of 135.32 degrees. We find that the angle between the front direction and a road with cars parked on the side is smaller. Therefore, the final answer is A. a road with cars parked on the side. | A. a road with cars parked on the side. | multi_object_facing | 002223b533d730f5.jpg |
00223b48f037a669_0cdc | Consider the real-world 3D locations of the objects. Is a white bird with long legs directly above a beach with a black background? | yes | no | null | null | [
{
"bbox_3d": [
0.7,
2,
10
],
"label": "a white bird with long legs"
},
{
"bbox_3d": [
-0.4,
0.2,
6.1
],
"label": "a beach with a black background"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a white bird with long legs and a beach with a black background. Then we compute the vector pointing from a beach with a black background to a white bird with long legs, as well as the up direction of a beach with a black background. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a white bird with long legs is directly above a beach with a black background. Otherwise, then a white bird with long legs is not directly above a beach with a black background. The 3D location of a white bird with long legs is (0.7, 2.0, 10.0). The 3D location of a beach with a black background is (-0.4, 0.2, 6.1). The vector from a beach with a black background to a white bird with long legs is hence (1.1, 1.8, 3.8). The up direction of a beach with a black background is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.41, corresponding to an angle of 65 degrees. The angle between the vector and the up direction is large, meaning that a white bird with long legs is not directly above a beach with a black background. Therefore, the answer is B. no. | B. no. | location_above | 00223b48f037a669.jpg |
002250249e2eaa0a_2b59 | Consider the real-world 3D locations and orientations of the objects. Which side of a computer with a black screen is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
0.7,
-0.5,
0.6
],
"label": "a computer with a black screen"
}
] | [
{
"front_dir": [
-0.9,
-0.2,
0.3
],
"label": "a computer with a black screen",
"left_dir": [
0.4,
-0.4,
0.8
]
}
] | D | To solve this problem, we first estimate the 3D location of a computer with a black screen. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a computer with a black screen, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a computer with a black screen that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a computer with a black screen is (0.7, -0.5, 0.6). The vector from a computer with a black screen to camera is hence (-0.7, 0.5, -0.6). The left direction of a computer with a black screen is (0.4, -0.4, 0.8). The cosine similarity between the vector pointing to camera and the left direction is -0.92, corresponding to an angle of 156.64 degrees. Thus the angle between the vector pointing to camera and the right direction is 23.36 degrees. The front direction of a computer with a black screen is (-0.9, -0.2, 0.3). The cosine similarity between the vector pointing to camera and the front direction is 0.34, corresponding to an angle of 70.31 degrees. Thus the angle between the vector pointing to camera and the back direction is 109.69 degrees. Among the four directions, the smallest angle is the right direction, with an angle of 23.36 degrees. Thus the right side of a computer with a black screen is facing the camera. Therefore, the final answer is D. right. | D. right. | orientation_viewpoint | 002250249e2eaa0a.jpg |
0022939df6108a93_6d84 | Consider the real-world 3D locations of the objects. Which is closer to a red hat with a white emblem, a white pillar with a sign on it or a pair of black furry shoes? | a white pillar with a sign on it | a pair of black furry shoes | null | null | [
{
"bbox_3d": [
0.1,
2.6,
4.1
],
"label": "a red hat with a white emblem"
},
{
"bbox_3d": [
0.8,
1.3,
3.7
],
"label": "a white pillar with a sign on it"
},
{
"bbox_3d": [
0,
0.2,
3.6
],
"label": "a pair of black furry shoes"
}
] | [] | A | To solve this problem, we first detect the 3D location of a red hat with a white emblem, a white pillar with a sign on it, and a pair of black furry shoes. Then we compute the L2 distances between a red hat with a white emblem and a white pillar with a sign on it, and between a red hat with a white emblem and a pair of black furry shoes. The object that is closer to a red hat with a white emblem is the one with a smaller distance. The 3D location of a red hat with a white emblem is (0.1, 2.6, 4.1). The 3D location of a white pillar with a sign on it is (0.8, 1.3, 3.7). The 3D location of a pair of black furry shoes is (-0.0, 0.2, 3.6). The L2 distance between a red hat with a white emblem and a white pillar with a sign on it is 1.4851315751783194. The L2 distance between a red hat with a white emblem and a pair of black furry shoes is 2.4174352312693665. Between the two distances, the distance between a red hat with a white emblem and a white pillar with a sign on it is smaller. Therefore, the final answer is A. a white pillar with a sign on it. | A. a white pillar with a sign on it. | multi_object_closer_to | 0022939df6108a93.jpg |
0022939df6108a93_20ad | Consider the real-world 3D locations of the objects. Is a man in a white dress directly underneath a man in a uniform stands in front of a stone building? | yes | no | null | null | [
{
"bbox_3d": [
1.3,
3.7,
11
],
"label": "a man in a uniform stands in front of a stone building"
},
{
"bbox_3d": [
0,
0.7,
4.1
],
"label": "a man in a white dress"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a man in a uniform stands in front of a stone building and a man in a white dress. Then we compute the vector pointing from a man in a white dress to a man in a uniform stands in front of a stone building, as well as the up direction of a man in a white dress. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a man in a uniform stands in front of a stone building is directly above a man in a white dress. Otherwise, then a man in a uniform stands in front of a stone building is not directly above a man in a white dress. To solve the question, we first determine if a man in a uniform stands in front of a stone building is directly above a man in a white dress. The 3D location of a man in a uniform stands in front of a stone building is (1.3, 3.7, 11.0). The 3D location of a man in a white dress is (-0.0, 0.7, 4.1). The vector from a man in a white dress to a man in a uniform stands in front of a stone building is hence (1.3, 3.0, 6.9). The up direction of a man in a white dress is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.39, corresponding to an angle of 67 degrees. The angle between the vector and the up direction is large, meaning that a man in a uniform stands in front of a stone building is not directly above a man in a white dress. In other words, a man in a white dress is not directly underneath a man in a uniform stands in front of a stone building. Therefore, the answer is B. no. | B. no. | location_above | 0022939df6108a93.jpg |
0022a423745512f2_6479 | Consider the real-world 3D locations and orientations of the objects. Which side of a brick pillar with a statue on top is facing the camera? | front | left | back | right | [
{
"bbox_3d": [
-2.7,
2.8,
9.9
],
"label": "a brick pillar with a statue on top"
}
] | [
{
"front_dir": [
-0.9,
-0.2,
-0.3
],
"label": "a brick pillar with a statue on top",
"left_dir": [
-0.3,
0.5,
0.8
]
}
] | D | To solve this problem, we first estimate the 3D location of a brick pillar with a statue on top. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a brick pillar with a statue on top, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a brick pillar with a statue on top that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a brick pillar with a statue on top is (-2.7, 2.8, 9.9). The vector from a brick pillar with a statue on top to camera is hence (2.7, -2.8, -9.9). The left direction of a brick pillar with a statue on top is (-0.3, 0.5, 0.8). The cosine similarity between the vector pointing to camera and the left direction is -0.97, corresponding to an angle of 166.98 degrees. Thus the angle between the vector pointing to camera and the right direction is 13.02 degrees. The front direction of a brick pillar with a statue on top is (-0.9, -0.2, -0.3). The cosine similarity between the vector pointing to camera and the front direction is 0.05, corresponding to an angle of 87.20 degrees. Thus the angle between the vector pointing to camera and the back direction is 92.80 degrees. Among the four directions, the smallest angle is the right direction, with an angle of 13.02 degrees. Thus the right side of a brick pillar with a statue on top is facing the camera. Therefore, the final answer is D. right. | D. right. | orientation_viewpoint | 0022a423745512f2.jpg |
0023a7e876e7cb6b_2616 | Consider the real-world 3D locations of the objects. Is a white plate with a white bowl directly underneath a white burrito with rice and vegetables? | yes | no | null | null | [
{
"bbox_3d": [
-0.1,
0.3,
0.5
],
"label": "a white burrito with rice and vegetables"
},
{
"bbox_3d": [
-0.1,
0.2,
0.3
],
"label": "a white plate with a white bowl"
}
] | [] | B | To solve this problem, we first determine the 3D locations of a white burrito with rice and vegetables and a white plate with a white bowl. Then we compute the vector pointing from a white plate with a white bowl to a white burrito with rice and vegetables, as well as the up direction of a white plate with a white bowl. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a white burrito with rice and vegetables is directly above a white plate with a white bowl. Otherwise, then a white burrito with rice and vegetables is not directly above a white plate with a white bowl. To solve the question, we first determine if a white burrito with rice and vegetables is directly above a white plate with a white bowl. The 3D location of a white burrito with rice and vegetables is (-0.1, 0.3, 0.5). The 3D location of a white plate with a white bowl is (-0.1, 0.2, 0.3). The vector from a white plate with a white bowl to a white burrito with rice and vegetables is hence (0.0, 0.1, 0.1). The up direction of a white plate with a white bowl is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.73, corresponding to an angle of 43 degrees. The angle between the vector and the up direction is large, meaning that a white burrito with rice and vegetables is not directly above a white plate with a white bowl. In other words, a white plate with a white bowl is not directly underneath a white burrito with rice and vegetables. Therefore, the answer is B. no. | B. no. | location_above | 0023a7e876e7cb6b.jpg |
002433c05e0ef11f_4169 | Consider the real-world 3D locations of the objects. Which is closer to a person is sitting on a concrete ledge, a bridge over water or a man wearing a black jacket? | a bridge over water | a man wearing a black jacket | null | null | [
{
"bbox_3d": [
-5.3,
1.5,
79.5
],
"label": "a person is sitting on a concrete ledge"
},
{
"bbox_3d": [
-13.8,
4.4,
136.9
],
"label": "a bridge over water"
},
{
"bbox_3d": [
6.1,
5.4,
42.4
],
"label": "a man wearing a black jacket"
}
] | [] | B | To solve this problem, we first detect the 3D location of a person is sitting on a concrete ledge, a bridge over water, and a man wearing a black jacket. Then we compute the L2 distances between a person is sitting on a concrete ledge and a bridge over water, and between a person is sitting on a concrete ledge and a man wearing a black jacket. The object that is closer to a person is sitting on a concrete ledge is the one with a smaller distance. The 3D location of a person is sitting on a concrete ledge is (-5.3, 1.5, 79.5). The 3D location of a bridge over water is (-13.8, 4.4, 136.9). The 3D location of a man wearing a black jacket is (6.1, 5.4, 42.4). The L2 distance between a person is sitting on a concrete ledge and a bridge over water is 58.11855786035807. The L2 distance between a person is sitting on a concrete ledge and a man wearing a black jacket is 39.0350742868004. Between the two distances, the distance between a person is sitting on a concrete ledge and a man wearing a black jacket is smaller. Therefore, the final answer is B. a man wearing a black jacket. | B. a man wearing a black jacket. | multi_object_closer_to | 002433c05e0ef11f.jpg |
0024a6207b572a5b_fd26 | Consider the real-world 3D locations and orientations of the objects. Which object is a stone building with a wooden balcony facing towards, a wooden balustrade with a bird perched on it or the a gravel alley? | a wooden balustrade with a bird perched on it | a gravel alley | null | null | [
{
"bbox_3d": [
-0.9,
4.2,
12.5
],
"label": "a stone building with a wooden balcony"
},
{
"bbox_3d": [
-3,
4.5,
13.7
],
"label": "a wooden balustrade with a bird perched on it"
},
{
"bbox_3d": [
-2.7,
0.1,
8.6
],
"label": "a gravel alley"
}
] | [
{
"front_dir": [
-0.5,
-0.2,
-0.9
],
"label": "a stone building with a wooden balcony",
"left_dir": [
-0.9,
0.2,
0.4
]
}
] | B | To solve this problem, we first detect the 3D location of a stone building with a wooden balcony, a wooden balustrade with a bird perched on it, and a gravel alley. Then we compute the cosine similarities between the front direction of a stone building with a wooden balcony and the vectors from a stone building with a wooden balcony to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a stone building with a wooden balcony is facing towards. The 3D location of a stone building with a wooden balcony is (-0.9, 4.2, 12.5). The 3D location of a wooden balustrade with a bird perched on it is (-3.0, 4.5, 13.7). The 3D location of a gravel alley is (-2.7, 0.1, 8.6). The front direction of a stone building with a wooden balcony is (-0.5, -0.2, -0.9). First we consider if a stone building with a wooden balcony is facing towards the a wooden balustrade with a bird perched on it. The vector from a stone building with a wooden balcony to a wooden balustrade with a bird perched on it is (-2.1, 0.3, 1.1). The cosine similarity between the front direction and the vector is -0.03, corresponding to an angle of 91.78 degrees. First we consider if a stone building with a wooden balcony is facing towards the a gravel alley. The vector from a stone building with a wooden balcony to a gravel alley is (-1.7, -4.1, -4.0). The cosine similarity between the front direction and the vector is 0.84, corresponding to an angle of 33.34 degrees. We find that the angle between the front direction and a gravel alley is smaller. Therefore, the final answer is B. a gravel alley. | B. a gravel alley. | multi_object_facing | 0024a6207b572a5b.jpg |
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 42