Dataset · omni-research
Short video detailed description benchmark
omni-research/DREAM-1K
A set of 1,000 short videos with detailed manual annotation of events, actions, and movements.
What's inside
Contains 1,000 ten-second clips from five sources: feature and animated films, stock footage, long YouTube videos, and TikTok-style short clips. Each video features detailed manual text annotations, with tabular data also included in the metadata.
What you can do with it
- Testing video models for accuracy in describing complex actions and events
- Evaluating subtitle generation and video content summarization quality
- Size
- 1K–10K rows
- Downloads this month
- 876
- Likes
- 35
- Access
- on request on Hugging Face
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("omni-research/DREAM-1K")