I have two questions about face detection model.
1. Is face detection model allowed in this contest? Since the statement says we can only use ONE DL model, so I would like to know about whether we can use face detection model.
2. If we can use face detection model, does it count on use extra data? As the face detection model use other face detection dataset to train the model.
Thanks for your answer.
Regards.Posted by: CV_newbie @ April 26, 2021, 11:30 p.m.
Yes, you can use a face detection model to detect the face. But you cannot use the extra data to train a model for face anti-spoofing.
Actually, we have used the face detector to detect faces in the HiFiMask dataset. Please read our paper (the data pre-processing of the Appendix part): https://arxiv.org/abs/2104.06148. In this paper, we also provide the labels of each specific attack material.
Thanks for your answer.Posted by: CV_newbie @ April 27, 2021, 12:18 a.m.
Another question, can we use face keypoint model or only allow to use face bounding box detection model?Posted by: CV_newbie @ May 7, 2021, 1:58 p.m.
The same problem.
Can we use the face detection model or the bounding box in the phase 2 test dataset?