Paul comments

Results 8 comments of


                                            Paul

which loss is lifted structure loss?

Thank you for your reply. I will keep focusing on the project. I just want to try lifted structure loss before in my dataset.

Will different image width and height in training set cause errors?

Thank you for your reply. I resize my training set to the same size but when run the code, I got the error message "ValueError: Dimension 1 in both shapes...

Will different image width and height in training set cause errors?

Thank you so much. I got everything done.

Will different image width and height in training set cause errors?

when I set --A_channels 3 --B_channels 1, I got message "Cannot feed value of shape (1, 256, 256, 3) for Tensor u'real_B:0', which has shape '(1, 256, 256, 1)'",so what...

About Evaluation

Yes, but what is "all shot "meaning？

Not getting a good accuracy

@VectorYYYY You mean batchsize 256 for every GPU or total batchsize 256 for 4 GPUs?

Not getting a good accuracy

1080ti can only set batchsize to 64 and I set 4 gpus for training. But I found loss around 2.1 cannot decrease and the model top1 accuracy is around 53%

My results always return 1.

@cvtower I find predict_class always outputs 1 whatever the predict_proba's output is (0, 1).