tipi2v commited on
Commit
c7f937d
·
verified ·
1 Parent(s): 291c7db

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +202 -0
README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: cc-by-nc-4.0
5
+ size_categories:
6
+ - 1M<n<10M
7
+ task_categories:
8
+ - image-to-video
9
+ - text-to-video
10
+ dataset_info:
11
+ features:
12
+ - name: UUID
13
+ dtype: string
14
+ - name: Text_Prompt
15
+ dtype: string
16
+ - name: Image_Prompt
17
+ dtype: image
18
+ - name: Subject
19
+ dtype: string
20
+ - name: Timestamp
21
+ dtype: string
22
+ - name: Text_NSFW
23
+ dtype: float32
24
+ - name: Image_NSFW
25
+ dtype: string
26
+ splits:
27
+ - name: Full
28
+ num_bytes: 13440652664.125
29
+ num_examples: 1701935
30
+ - name: Subset
31
+ num_bytes: 790710630
32
+ num_examples: 100000
33
+ - name: Eval
34
+ num_bytes: 78258893
35
+ num_examples: 10000
36
+ download_size: 27500759907
37
+ dataset_size: 27750274851.25
38
+ configs:
39
+ - config_name: default
40
+ data_files:
41
+ - split: Full
42
+ path: data/Full-*
43
+ - split: Subset
44
+ path: data/Subset-*
45
+ - split: Eval
46
+ path: data/Eval-*
47
+ tags:
48
+ - prompt
49
+ - image-to-video
50
+ - text-to-video
51
+ - visual-generation
52
+ - video-generation
53
+ pretty_name: TIP-I2V
54
+ ---
55
+
56
+ # Summary
57
+ This is the dataset proposed in our paper **TIP-I2V: A Million-Scale Real Prompt-Gallery Dataset for Image-to-Video Diffusion Models**.
58
+
59
+ TIP-I2V is the first dataset comprising over 1.70 million unique user-provided text and image prompts. Besides the prompts, TIP-I2V also includes videos generated by five state-of-the-art image-to-video models (Pika, Stable Video Diffusion, Open-Sora, I2VGen-XL, and CogVideoX-5B). The TIP-I2V contributes to the development of better and safer image-to-video models.
60
+
61
+ <p align="center">
62
+ <img src="https://huggingface.co/datasets/TIP-I2V/TIP-I2V/resolve/main/assets/teasor.png" width="1000">
63
+ </p>
64
+
65
+ # Datapoint
66
+ <p align="center">
67
+ <img src="https://huggingface.co/datasets/TIP-I2V/TIP-I2V/resolve/main/assets/datapoint.png" width="1000">
68
+ </p>
69
+
70
+ # Statistics
71
+ <p align="center">
72
+ <img src="https://huggingface.co/datasets/TIP-I2V/TIP-I2V/resolve/main/assets/stat.png" width="1000">
73
+ </p>
74
+
75
+ # Download
76
+
77
+ For users in mainland China, try setting `export HF_ENDPOINT=https://hf-mirror.com` to successfully download the weights.
78
+ ## Download the text and (compressed) image prompts with related information
79
+
80
+ ```python
81
+ # Full (text and compressed image) prompts: ~13.4G
82
+ from datasets import load_dataset
83
+ ds = load_dataset("TIP-I2V/TIP-I2V", split='Full', streaming=True)
84
+
85
+ # Convert to Pandas format (it may be slow)
86
+ import pandas as pd
87
+ df = pd.DataFrame(ds)
88
+ ```
89
+
90
+
91
+ ```python
92
+ # 100k subset (text and compressed image) prompts: ~0.8G
93
+ from datasets import load_dataset
94
+ ds = load_dataset("TIP-I2V/TIP-I2V", split='Subset', streaming=True)
95
+
96
+ # Convert to Pandas format (it may be slow)
97
+ import pandas as pd
98
+ df = pd.DataFrame(ds)
99
+ ```
100
+
101
+ ```python
102
+ # 10k TIP-Eval (text and compressed image) prompts: ~0.08G
103
+ from datasets import load_dataset
104
+ ds = load_dataset("TIP-I2V/TIP-I2V", split='Eval', streaming=True)
105
+
106
+ # Convert to Pandas format (it may be slow)
107
+ import pandas as pd
108
+ df = pd.DataFrame(ds)
109
+ ```
110
+
111
+ ## Download the embeddings for text and image prompts
112
+
113
+ ```python
114
+ # Embeddings for full text prompts (~21G) and image prompts (~3.5G)
115
+ from huggingface_hub import hf_hub_download
116
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Full_Text_Embedding.parquet", repo_type="dataset")
117
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Full_Image_Embedding.parquet", repo_type="dataset")
118
+ ```
119
+
120
+ ```python
121
+ # Embeddings for 100k subset text prompts (~1.2G) and image prompts (~0.2G)
122
+ from huggingface_hub import hf_hub_download
123
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Subset_Text_Embedding.parquet", repo_type="dataset")
124
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Subset_Image_Embedding.parquet", repo_type="dataset")
125
+ ```
126
+
127
+ ```python
128
+ # Embeddings for 10k TIP-Eval text prompts (~0.1G) and image prompts (~0.02G)
129
+ from huggingface_hub import hf_hub_download
130
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Eval_Text_Embedding.parquet", repo_type="dataset")
131
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="Embedding/Eval_Image_Embedding.parquet", repo_type="dataset")
132
+ ```
133
+
134
+ ## Download uncompressed image prompts
135
+
136
+ ```python
137
+ # Full uncompressed image prompts: ~1T
138
+ from huggingface_hub import hf_hub_download
139
+ for i in range(1,52):
140
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="image_prompt_tar/image_prompt_%d.tar"%i, repo_type="dataset")
141
+ ```
142
+
143
+ ```python
144
+ # 100k subset uncompressed image prompts: ~69.6G
145
+ from huggingface_hub import hf_hub_download
146
+ for i in range(1,3):
147
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="sub_image_prompt_tar/sub_image_prompt_%d.tar"%i, repo_type="dataset")
148
+ ```
149
+
150
+ ```python
151
+ # 10k TIP-Eval uncompressed image prompts: ~6.5G
152
+ from huggingface_hub import hf_hub_download
153
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_image_prompt_tar/eval_image_prompt.tar", repo_type="dataset")
154
+ ```
155
+
156
+ ## Download generated videos
157
+
158
+ ```python
159
+ # Full videos generated by Pika: ~1T
160
+ from huggingface_hub import hf_hub_download
161
+ for i in range(1,52):
162
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="pika_videos_tar/pika_videos_%d.tar"%i, repo_type="dataset")
163
+ ```
164
+
165
+ ```python
166
+ # 100k subset videos generated by Pika (~57.6G), Stable Video Diffusion (~38.9G), Open-Sora (~47.2G), I2VGen-XL (~54.4G), and CogVideoX-5B (~36.7G)
167
+ from huggingface_hub import hf_hub_download
168
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/pika_videos_subset_1.tar", repo_type="dataset")
169
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/pika_videos_subset_2.tar", repo_type="dataset")
170
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/svd_videos_subset.tar", repo_type="dataset")
171
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/opensora_videos_subset.tar", repo_type="dataset")
172
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/i2vgenxl_videos_subset_1.tar", repo_type="dataset")
173
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/i2vgenxl_videos_subset_2.tar", repo_type="dataset")
174
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="subset_videos_tar/cog_videos_subset.tar", repo_type="dataset")
175
+ ```
176
+
177
+ ```python
178
+ # 10k TIP-Eval videos generated by Pika (~5.8G), Stable Video Diffusion (~3.9G), Open-Sora (~4.7G), I2VGen-XL (~5.4G), and CogVideoX-5B (~3.6G)
179
+ from huggingface_hub import hf_hub_download
180
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_videos_tar/pika_videos_eval.tar", repo_type="dataset")
181
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_videos_tar/svd_videos_eval.tar", repo_type="dataset")
182
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_videos_tar/opensora_videos_eval.tar", repo_type="dataset")
183
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_videos_tar/i2vgenxl_videos_eval.tar", repo_type="dataset")
184
+ hf_hub_download(repo_id="TIP-I2V/TIP-I2V", filename="eval_videos_tar/cog_videos_eval.tar", repo_type="dataset")
185
+ ```
186
+
187
+ # Comparison with VidProM and DiffusionDB
188
+ <p align="center">
189
+ <img src="https://huggingface.co/datasets/TIP-I2V/TIP-I2V/resolve/main/assets/table.png" width="1000">
190
+ </p>
191
+ <p align="center">
192
+ <img src="https://huggingface.co/datasets/TIP-I2V/TIP-I2V/resolve/main/assets/comparison.png" width="1000">
193
+ </p>
194
+
195
+ Click the [WizMap (TIP-I2V VS VidProM)](https://poloclub.github.io/wizmap/?dataURL=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FTIP-I2V%2FTIP-I2V%2Fresolve%2Fmain%2Ftip-i2v-visualize%2Fdata_tip-i2v_vidprom.ndjson&gridURL=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FTIP-I2V%2FTIP-I2V%2Fresolve%2Fmain%2Ftip-i2v-visualize%2Fgrid_tip-i2v_vidprom.json) and [WizMap (TIP-I2V VS DiffusionDB)](https://poloclub.github.io/wizmap/?dataURL=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FTIP-I2V%2FTIP-I2V%2Fresolve%2Fmain%2Ftip-i2v-visualize%2Fdata_tip-i2v_diffusiondb.ndjson&gridURL=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FTIP-I2V%2FTIP-I2V%2Fresolve%2Fmain%2Ftip-i2v-visualize%2Fgrid_tip-i2v_diffusiondb.json)
196
+ (wait for 5 seconds) for an interactive visualization of our 1.70 million prompts.
197
+
198
+
199
+ # License
200
+
201
+ The prompts and videos in our TIP-I2V are licensed under the [CC BY-NC 4.0 license](https://creativecommons.org/licenses/by-nc/4.0/deed.en).
202
+