35,666 high-quality video clips across 83 facial attribute classes and 22 action categories. The collection features 15,653 unique identities with high-resolution temporal data.
Use Cases
- Train video-based facial attribute recognition models using the 83 attribute labels
- Develop emotion recognition systems leveraging the 15 emotion categories
- Perform identity-consistent face synthesis or editing using the 15,653 unique identity labels
Strengths
- 35,666 video clips in high-definition resolution
- 83 distinct facial attributes and 22 action categories
- 15,653 unique celebrity identities