Dataset · ego2act
Ego2act benchmark for video generation
ego2act/ego2act-bench
A benchmark for evaluating goal-directed manipulations in video generation using egocentric frames.
What's inside
The dataset contains under 1K entries with video, images, and text in English. Includes tasks and reference records of human manipulations to evaluate models by initial frame and text goal.
What you can do with it
- Evaluating and testing video generation models (image-to-video)
- Research in video classification and egocentric vision
- Tasks
- Image → video, Video classification
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 533
- Likes
- 4
How to load it
from datasets import load_dataset
ds = load_dataset("ego2act/ego2act-bench")