anhttran/3dmm_cnn

about the asymmetric Euclidean loss

Opened this issue · 0 comments

Thank you for sharing your code and your great jobs about 3dmm_basic and this. When I use the loss function on another network, like mobilenet, to train 3dmm parameters, it does not perform so well. I don't know what went wrong. I have doubts about how γmax is calculated?