Pytorch implementation for Learning Symmetry Consistent Deep CNNs for Face Completion
Overview of our SymmFCNet. Red, green and blue lines represent the pixel-wise correspondence between the input and the flip image.
- Red: missing pixels (input) to non-occluded pixels (flip);
- Green: missing pixels (input) to missing pixels (flip);
- Blue: remaining pixels (input) to remaining pixels (flip).
Pytorch source code and models will come soon.