NVIDIA / vid2vid

Pytorch implementation of our method for high-resolution (e.g. 2048x1024) photorealistic video-to-video translation.
Other
8.57k stars 1.2k forks source link

Faces not being centercropped at inference #156

Closed gabewilliam closed 4 years ago

gabewilliam commented 4 years ago

I am running tests using the pretrained face model and provided datasets, though for most of them the faces aren't centercropped. In a lot of examples, the face isn't even IN the cropped area.

I've plotted the images with the detected landmarks, and they line up perfectly, so it doesn't seem that the landmarks are off.

Fake: fake_B_00030 Edgemap: real_A_00030 Real: 00029

Does anyone know what's causing this?