22,176 facial images featuring 11-class pixel-level semantic segmentation masks and 106-point facial landmarks. The collection includes significant variations in pose, expression, and occlusion to facilitate research in fine-grained face parsing.
Use Cases
- Train deep learning models for semantic segmentation using the 11-class pixel-level labels
- Develop landmark localization algorithms using the 106-point facial coordinate data
- Perform multi-task learning to improve boundary accuracy by combining parsing masks and landmark features
Strengths
- 22,176 high-resolution images with corresponding pixel-level annotations
- 11 distinct semantic categories including hair, facial skin, and specific sensory organs
- 106-point facial landmark coordinates provided for every image in the dataset
- Data splits organized into 18,176 training, 2,000 validation, and 2,000 testing samples