Dataset · XiangpengYang
Video editing dataset for VideoGrain
XiangpengYang/VideoGrain-dataset
English image and data collection from the VideoGrain paper (ICLR 2025) for multi-level video editing tasks.
What's inside
The dataset was prepared by the authors of the VideoGrain research paper (accepted at ICLR 2025), which focuses on zero-shot video editing with spatio-temporal attention control. The metadata specifies image modality, text-to-video and text-to-image tasks, English language, and a size ranging from 1 to 10 thousand records.
What you can do with it
- Testing and comparing zero-shot object and class editing methods in videos
- Evaluating style transfer and local modification algorithms in generative video models
- Reproducing experiments from the VideoGrain paper
- Tasks
- Text → video, Text → image
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 40.3K
- Likes
- 1
How to load it
from datasets import load_dataset
ds = load_dataset("XiangpengYang/VideoGrain-dataset")