daojishigailvlun/HeadPose

HeadPose

1. Face detection
2. Datasets
3. How to use our model
4. Reference

Face detection

We used the face detect method proposed by Chen et al. We will give the bounding box in the list. You can also use Microsoft face API to get the same result if you want to use our method in your own dataset.

Datasets

• 300W, AFLW2000
• BIWI
• SASE

Train

To train on 300W and test on AFLW2000 with bounding box margin = 0.5
python pose.py 0.5

Reference

If you find the work useful in your research please consider citing:

@article{shao2019FG
  author    = {Mingzhen Shao and Zhun Sun and Mete Ozay and Takayuki Okatani},
  title     = {Improving Head Pose Estimation with a Combined Loss and Bounding Box Margin Adjustment},
  booktitle = {The 14th IEEE International Conference on Automatic Face and Gesture Recognition (FG2019)}
  year      = {2019}