Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,86 @@
|
|
1 |
-
---
|
2 |
-
license: cc-by-4.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-4.0
|
3 |
+
task_categories:
|
4 |
+
- text-to-video
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
tags:
|
8 |
+
- text-to-video
|
9 |
+
- Video Generative Model Training
|
10 |
+
- Text-to-Video Diffusion Model Training
|
11 |
+
- prompts
|
12 |
+
pretty_name: OpenVid-1M-mapping
|
13 |
+
size_categories:
|
14 |
+
- 1M<n<10M
|
15 |
+
---
|
16 |
+
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
# Summary
|
21 |
+
This is the dataset proposed in the paper "[**OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation**](https://arxiv.org/abs/2407.02371)".
|
22 |
+
OpenVid-1M is a high-quality text-to-video dataset designed for research institutions to enhance video quality, featuring high aesthetics, clarity, and resolution. It can be used for direct training or as a quality tuning complement to other video datasets.
|
23 |
+
|
24 |
+
**New Feature**: Video-ZIP mapping files now available for efficient video lookup (see [Dataset Structure](#dataset-structure) and [Extended Usage](#extended-usage)).
|
25 |
+
|
26 |
+
# Dataset Structure
|
27 |
+
```
|
28 |
+
DATA_PATH
|
29 |
+
└── video_mappings (NEW!)
|
30 |
+
├── OpenVid_part0.csv
|
31 |
+
├── OpenVid_part1.csv
|
32 |
+
├── ...
|
33 |
+
└── OpenVid_part185.csv
|
34 |
+
├── README.md
|
35 |
+
└── ...
|
36 |
+
```
|
37 |
+
|
38 |
+
# Key Updates
|
39 |
+
1. **Video-ZIP Mapping**
|
40 |
+
We provide 186 mapping files (`OpenVid_part*.csv`) containing:
|
41 |
+
```csv
|
42 |
+
zip_file,video,video_path
|
43 |
+
OpenVid_part0.zip,celebv_--Jiv5iYqT8_0.mp4,OpenVid_part1/celebv_--Jiv5iYqT8_0.mp4
|
44 |
+
OpenVid_part0.zip,celebv_--Jiv5iYqT8_2.mp4,OpenVid_part1/celebv_--Jiv5iYqT8_2.mp4
|
45 |
+
OpenVid_part0.zip,celebv_--QCZKgJt6o_0.mp4,OpenVid_part1/celebv_--QCZKgJt6o_0.mp4
|
46 |
+
OpenVid_part0.zip,celebv_--oCWVOBuvA_0.mp4,OpenVid_part1/celebv_--oCWVOBuvA_0.mp4
|
47 |
+
...
|
48 |
+
```
|
49 |
+
|
50 |
+
- `zip_file`: Source ZIP archive name
|
51 |
+
- `video`: Video filename
|
52 |
+
- `video_path`: Relative path within the ZIP file
|
53 |
+
|
54 |
+
2. **Mapping Features**
|
55 |
+
- Enables direct video lookup without full ZIP extraction
|
56 |
+
- Contains cross-references between ZIP files and video paths
|
57 |
+
- Split into multiple files for parallel processing
|
58 |
+
|
59 |
+
# Extended Usage
|
60 |
+
## 1. Video Mapping Access
|
61 |
+
```python
|
62 |
+
import pandas as pd
|
63 |
+
|
64 |
+
# Load all mapping files
|
65 |
+
mappings = [pd.read_csv(f"video_mappings/OpenVid_part{i}.csv")
|
66 |
+
for i in range(186)]
|
67 |
+
full_mapping = pd.concat(mappings)
|
68 |
+
|
69 |
+
# Find ZIP location for specific video
|
70 |
+
video_name = "celebv_--Jiv5iYqT8_0.mp4"
|
71 |
+
result = full_mapping[full_mapping.video == video_name]
|
72 |
+
print(f"Video {video_name} is in {result.zip_file.values[0]} at path {result.video_path.values[0]}")
|
73 |
+
```
|
74 |
+
|
75 |
+
## 2. Advanced ZIP Handling
|
76 |
+
### Extraction Specific Video from Zip_file
|
77 |
+
```bash
|
78 |
+
# Extract specific video from ZIP without full decompression
|
79 |
+
unzip OpenVid_part0.zip "OpenVid_part1/celebv_--Jiv5iYqT8_0.mp4" -d target_folder
|
80 |
+
# Unzip files (-j flattens directory structure)
|
81 |
+
```
|
82 |
+
|
83 |
+
|
84 |
+
|
85 |
+
# Acknowledgments
|
86 |
+
The video-ZIP mapping system was developed to enhance dataset accessibility while maintaining original directory structures. Special thanks to the original dataset contributors.
|