Datasets:
Tasks:
Text-to-Video
Size:
1K - 10K
ArXiv:
Tags:
video
text-to-video
video-generation
reinforcement-learning
reward-optimization
synthetic-data
License:
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -76,8 +76,6 @@ GraftSet-Video is the external video exemplar dataset introduced in **GraftOOD:
|
|
| 76 |
|
| 77 |
## Metadata schema
|
| 78 |
|
| 79 |
-
`metadata.jsonl` : one JSON object per video.
|
| 80 |
-
|
| 81 |
| Field | Type | Description |
|
| 82 |
|---|---|---|
|
| 83 |
| `file_name` | string | Video path relative to the directory containing `metadata.jsonl`, such as `videos/wan2_2/1_wan2_2.mp4`. |
|
|
@@ -102,8 +100,6 @@ Example record, formatted across multiple lines for readability:
|
|
| 102 |
|
| 103 |
## Loading the dataset
|
| 104 |
|
| 105 |
-
### Read a local copy without decoding videos
|
| 106 |
-
|
| 107 |
Run the following from the dataset repository root. It uses only the Python standard library:
|
| 108 |
|
| 109 |
```python
|
|
|
|
| 76 |
|
| 77 |
## Metadata schema
|
| 78 |
|
|
|
|
|
|
|
| 79 |
| Field | Type | Description |
|
| 80 |
|---|---|---|
|
| 81 |
| `file_name` | string | Video path relative to the directory containing `metadata.jsonl`, such as `videos/wan2_2/1_wan2_2.mp4`. |
|
|
|
|
| 100 |
|
| 101 |
## Loading the dataset
|
| 102 |
|
|
|
|
|
|
|
| 103 |
Run the following from the dataset repository root. It uses only the Python standard library:
|
| 104 |
|
| 105 |
```python
|