A implementaion of depth generation based on [PRNet](https://github.com/YadiraF/PRNet), which was used in the paper ***Exploiting Temporal and Depth Information for Multi-frame Face Anti-spoofing***
The result generated from your code is different from PRNet souce code https://github.com/YadiraF/PRNet)
I just test trump.jpg, the result of depth image is difficult.
We add the depth generation based on a past version of PRNet, and haven't implemented the z-buffer, so if your test image contains minor/large-pose face, the unseen facial parts will influence the depth results.
The result generated from your code is different from PRNet souce code https://github.com/YadiraF/PRNet) I just test trump.jpg, the result of depth image is difficult.