search

cleardusk / 3DDFA

The PyTorch improved version of TPAMI 2017 paper: Face Alignment in Full Pose Range: A 3D Total Solution.
MIT License
3.61k stars 653 forks source link

How to estimate the quality of depth images generated #164

Open YokkaBear opened 4 years ago

YokkaBear commented 4 years ago

Hi @cleardusk , I have used main.py and the model phase1_wpdc_vdc.pth.tar to generate some depth images from my own human face dataset, but the resulting depth images do not look quite similar to the original faces and seem lose many details.

So, I wonder how I could measure the quality of the depth images generated, or if I ignore any scripts in this repo just to realize this function. Looking forward to your reply, much thanks.

cleardusk commented 4 years ago

The depth image generated may be called Pseudo-depth. If you want detailed depth, you may refer to the work for the detailed 3d face reconstruction, or use physical devices to obtain them.