|
|
--- |
|
|
license: mit |
|
|
pipeline_tag: keypoint-detection |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- mse |
|
|
datasets: |
|
|
- WFLW |
|
|
- AFLW |
|
|
- COFW |
|
|
- 300W |
|
|
--- |
|
|
### Model Description |
|
|
|
|
|
This model is designed for **facial landmark detection (face alignment)**, aiming to accurately localize landmarks under challenging conditions such as: |
|
|
|
|
|
- Large pose variations |
|
|
- Occlusion |
|
|
- Illumination changes |
|
|
|
|
|
It is based on a Transformer-based architecture with boundary-aware and fine-grained multi-task balancing mechanisms. |
|
|
|
|
|
The model achieves strong performance across multiple benchmark datasets. |
|
|
This model is associated with the following paper: https://www.huggingface.co/papers/2601.12863 |
|
|
For a detailed description of the methodology, experiments and analysis, please refer to the paper page above. |
|
|
|
