adepierre / caffe_cyclegan Goto Github PK

Hi, when you say code relies on caffe with modifications to solver.hpp, do you mean I have to make the changes to solver.hpp then re-Build/re-compile my whole installation of caffe?
thanks

Hi, @adepierre ~
I found a weird problem when training.
It is weird that the program is killed unexpectedly when iteration is 62250.
But nobody kills the program.
There are 753 image pairs for training.

Loss on real A data: 0.0568343 on generated A data: 0.0570172
Loss on real B data: 0.0519266 on generated B data: 0.0524207
Cyclic loss on A data: 0.229246
Cyclic loss on B data: 0.164716
I0712 23:00:19.389642 45804 sgd_solver.cpp:112] Iteration 62250, lr = 0.0002
I0712 23:00:19.398865 45804 sgd_solver.cpp:112] Iteration 62250, lr = 0.0002
I0712 23:00:19.401080 45804 sgd_solver.cpp:112] Iteration 62250, lr = 0.0002
I0712 23:00:19.401469 45804 sgd_solver.cpp:112] Iteration 62250, lr = 0.0002
Killed

Could you please answer this question?
Many thanks!!

As far as I understand, the output of the discriminator outputs a scalar which represents whether the input comes from natural image set or generated image set. But I find that your discriminator outputs a 32x32 tensor. Could you explain the reason of that? Thx

Hi, Caffe_CycleGAN.exe seems to crash (windows) on batch_size > 2.

Is there anything in the code that needs changing to increase batchsize?
Thanks

Hi, Can you share your trained model? Thank you!

No bugs found, and the solver prototxt is OK, any ideas？

Hi, @adepierre
I really appreciate your work~
But, i want to know how can i use it in Linux.

Could you please give me more details for running in Linux?
Many thanks!

Valentina

What should I do?