Spaces:
Runtime error
Runtime error
Old video removal mechanism.
Browse files- app.py +32 -2
- utils/video.py +32 -0
app.py
CHANGED
|
@@ -10,11 +10,34 @@ from tqdm import tqdm
|
|
| 10 |
from inference.models import YOLOWorld
|
| 11 |
|
| 12 |
from utils.efficient_sam import load, inference_with_boxes
|
| 13 |
-
from utils.video import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
MARKDOWN = """
|
| 16 |
# YOLO-World + EfficientSAM 🔥
|
| 17 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
This is a demo of zero-shot object detection and instance segmentation using
|
| 19 |
[YOLO-World](https://github.com/AILab-CVC/YOLO-World) and
|
| 20 |
[EfficientSAM](https://github.com/yformer/EfficientSAM).
|
|
@@ -35,6 +58,7 @@ RESULTS = "results"
|
|
| 35 |
|
| 36 |
IMAGE_EXAMPLES = [
|
| 37 |
['https://media.roboflow.com/dog.jpeg', 'dog, eye, nose, tongue, car', 0.005, 0.1, True, False, False],
|
|
|
|
| 38 |
]
|
| 39 |
VIDEO_EXAMPLES = [
|
| 40 |
['https://media.roboflow.com/supervision/video-examples/croissant-1280x720.mp4', 'croissant', 0.01, 0.2, False, False, False],
|
|
@@ -51,7 +75,7 @@ BOUNDING_BOX_ANNOTATOR = sv.BoundingBoxAnnotator()
|
|
| 51 |
MASK_ANNOTATOR = sv.MaskAnnotator()
|
| 52 |
LABEL_ANNOTATOR = sv.LabelAnnotator()
|
| 53 |
|
| 54 |
-
|
| 55 |
create_directory(directory_path=RESULTS)
|
| 56 |
|
| 57 |
|
|
@@ -89,6 +113,9 @@ def process_image(
|
|
| 89 |
with_confidence: bool = False,
|
| 90 |
with_class_agnostic_nms: bool = False,
|
| 91 |
) -> np.ndarray:
|
|
|
|
|
|
|
|
|
|
| 92 |
categories = process_categories(categories)
|
| 93 |
YOLO_WORLD_MODEL.set_classes(categories)
|
| 94 |
results = YOLO_WORLD_MODEL.infer(input_image, confidence=confidence_threshold)
|
|
@@ -124,6 +151,9 @@ def process_video(
|
|
| 124 |
with_class_agnostic_nms: bool = False,
|
| 125 |
progress=gr.Progress(track_tqdm=True)
|
| 126 |
) -> str:
|
|
|
|
|
|
|
|
|
|
| 127 |
categories = process_categories(categories)
|
| 128 |
YOLO_WORLD_MODEL.set_classes(categories)
|
| 129 |
video_info = sv.VideoInfo.from_video_path(input_video)
|
|
|
|
| 10 |
from inference.models import YOLOWorld
|
| 11 |
|
| 12 |
from utils.efficient_sam import load, inference_with_boxes
|
| 13 |
+
from utils.video import (
|
| 14 |
+
generate_file_name,
|
| 15 |
+
calculate_end_frame_index,
|
| 16 |
+
create_directory,
|
| 17 |
+
remove_files_older_than
|
| 18 |
+
)
|
| 19 |
|
| 20 |
MARKDOWN = """
|
| 21 |
# YOLO-World + EfficientSAM 🔥
|
| 22 |
|
| 23 |
+
<div>
|
| 24 |
+
<a href="https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/zero-shot-object-detection-with-yolo-world.ipynb">
|
| 25 |
+
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Colab" style="display:inline-block;">
|
| 26 |
+
</a>
|
| 27 |
+
<a href="https://blog.roboflow.com/what-is-yolo-world/">
|
| 28 |
+
<img src="https://raw.githubusercontent.com/roboflow-ai/notebooks/main/assets/badges/roboflow-blogpost.svg" alt="Roboflow" style="display:inline-block;">
|
| 29 |
+
</a>
|
| 30 |
+
<a href="https://www.youtube.com/watch?v=X7gKBGVz4vs">
|
| 31 |
+
<img src="https://badges.aleen42.com/src/youtube.svg" alt="YouTube" style="display:inline-block;">
|
| 32 |
+
</a>
|
| 33 |
+
<a href="https://github.com/AILab-CVC/YOLO-World">
|
| 34 |
+
<img src="https://badges.aleen42.com/src/github.svg" alt="GitHub" style="display:inline-block;">
|
| 35 |
+
</a>
|
| 36 |
+
<a href="https://arxiv.org/abs/2401.17270">
|
| 37 |
+
<img src="https://img.shields.io/badge/arXiv-2401.17270-b31b1b.svg" alt="arXiv" style="display:inline-block;">
|
| 38 |
+
</a>
|
| 39 |
+
</div>
|
| 40 |
+
|
| 41 |
This is a demo of zero-shot object detection and instance segmentation using
|
| 42 |
[YOLO-World](https://github.com/AILab-CVC/YOLO-World) and
|
| 43 |
[EfficientSAM](https://github.com/yformer/EfficientSAM).
|
|
|
|
| 58 |
|
| 59 |
IMAGE_EXAMPLES = [
|
| 60 |
['https://media.roboflow.com/dog.jpeg', 'dog, eye, nose, tongue, car', 0.005, 0.1, True, False, False],
|
| 61 |
+
['https://media.roboflow.com/albert-4x.png', 'hand, hair', 0.005, 0.1, True, False, False],
|
| 62 |
]
|
| 63 |
VIDEO_EXAMPLES = [
|
| 64 |
['https://media.roboflow.com/supervision/video-examples/croissant-1280x720.mp4', 'croissant', 0.01, 0.2, False, False, False],
|
|
|
|
| 75 |
MASK_ANNOTATOR = sv.MaskAnnotator()
|
| 76 |
LABEL_ANNOTATOR = sv.LabelAnnotator()
|
| 77 |
|
| 78 |
+
# creating video results directory
|
| 79 |
create_directory(directory_path=RESULTS)
|
| 80 |
|
| 81 |
|
|
|
|
| 113 |
with_confidence: bool = False,
|
| 114 |
with_class_agnostic_nms: bool = False,
|
| 115 |
) -> np.ndarray:
|
| 116 |
+
# cleanup of old video files
|
| 117 |
+
remove_files_older_than(RESULTS, 30)
|
| 118 |
+
|
| 119 |
categories = process_categories(categories)
|
| 120 |
YOLO_WORLD_MODEL.set_classes(categories)
|
| 121 |
results = YOLO_WORLD_MODEL.infer(input_image, confidence=confidence_threshold)
|
|
|
|
| 151 |
with_class_agnostic_nms: bool = False,
|
| 152 |
progress=gr.Progress(track_tqdm=True)
|
| 153 |
) -> str:
|
| 154 |
+
# cleanup of old video files
|
| 155 |
+
remove_files_older_than(RESULTS, 30)
|
| 156 |
+
|
| 157 |
categories = process_categories(categories)
|
| 158 |
YOLO_WORLD_MODEL.set_classes(categories)
|
| 159 |
video_info = sv.VideoInfo.from_video_path(input_video)
|
utils/video.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
import os
|
| 2 |
import datetime
|
| 3 |
import uuid
|
|
|
|
| 4 |
|
| 5 |
import supervision as sv
|
| 6 |
|
|
@@ -14,6 +15,37 @@ def generate_file_name(extension="mp4"):
|
|
| 14 |
return f"{current_datetime}_{unique_id}.{extension}"
|
| 15 |
|
| 16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
def calculate_end_frame_index(source_video_path: str) -> int:
|
| 18 |
video_info = sv.VideoInfo.from_video_path(source_video_path)
|
| 19 |
return min(
|
|
|
|
| 1 |
import os
|
| 2 |
import datetime
|
| 3 |
import uuid
|
| 4 |
+
from typing import List
|
| 5 |
|
| 6 |
import supervision as sv
|
| 7 |
|
|
|
|
| 15 |
return f"{current_datetime}_{unique_id}.{extension}"
|
| 16 |
|
| 17 |
|
| 18 |
+
def list_files_older_than(directory: str, diff_minutes: int) -> List[str]:
|
| 19 |
+
diff_seconds = diff_minutes * 60
|
| 20 |
+
now = datetime.datetime.now()
|
| 21 |
+
older_files: List[str] = []
|
| 22 |
+
|
| 23 |
+
for filename in os.listdir(directory):
|
| 24 |
+
file_path = os.path.join(directory, filename)
|
| 25 |
+
if os.path.isfile(file_path):
|
| 26 |
+
file_mod_time = os.path.getmtime(file_path)
|
| 27 |
+
file_mod_datetime = datetime.datetime.fromtimestamp(file_mod_time)
|
| 28 |
+
time_diff = now - file_mod_datetime
|
| 29 |
+
if time_diff.total_seconds() > diff_seconds:
|
| 30 |
+
older_files.append(file_path)
|
| 31 |
+
|
| 32 |
+
return older_files
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
def remove_files_older_than(directory: str, diff_minutes: int) -> None:
|
| 36 |
+
older_files = list_files_older_than(directory, diff_minutes)
|
| 37 |
+
file_count = len(older_files)
|
| 38 |
+
|
| 39 |
+
for file_path in older_files:
|
| 40 |
+
os.remove(file_path)
|
| 41 |
+
|
| 42 |
+
now = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
| 43 |
+
print(
|
| 44 |
+
f"[{now}] Removed {file_count} files older than {diff_minutes} minutes from "
|
| 45 |
+
f"'{directory}' directory."
|
| 46 |
+
)
|
| 47 |
+
|
| 48 |
+
|
| 49 |
def calculate_end_frame_index(source_video_path: str) -> int:
|
| 50 |
video_info = sv.VideoInfo.from_video_path(source_video_path)
|
| 51 |
return min(
|