ZYD
In the paper GAN,the loss funtion has a form of minmaxV(G,D),while in your code the loss is 'binary_crossentropy',I'm little confused about this,could you please explain this a little bit?
https://github.com/kuangliu/torchcv/blob/ca4480243c9680146bf8815253e465dd7a80da84/torchcv/models/ssd/net.py#L134 In this line why you permute loc_pred to (0, 2, 3, 1) and why you use contiguous? I am confused here
it seems that you train your model only 200 epochs from scratch. I found that not work, could you please post pretrained base vgg model and your learning details?