The End-to-End in code

[range428]

I carefully read your code and wanted to find the end-to-end learning process of “GAN and Det”, but I didn't find it. I think esrgan EESN FRCNN_ Model.py only implements segmentation training。

...
l_g_total.backward(retain_graph=True)  
self.optimizer_G.step()  
...
l_d_total.backward()
self.optimizer_D.step()
...
losses.backward()
self.optimizer_FRCNN.step()

I don't think this part realizes that the loss of detector and discriminator is back propagation to GAN Can you release the code for end-to-end training?

[KZGSK]

I have the same question.I don't know whether it is ok to add the loss of FRCNN to G.I wonder what you think.Looking forward to your reply！