is it place vgg_19.ckpt in the dictionary straightly?but keyerror:vgg19_1/vgg_19/conv2/conv2_2

[YueHan0525]

No description provided.

[avisekiit]

Hi. Could you solve this issue? I am facing a similar problem. I tried to train SRRESNET with MSE loss which looked fine but when I tried to initiate SRAGN phase with the vgg54 loss I also got a similar kind of error message.

[YueHan0525]

modified 74 rows "return layers[layer]" in srgan.py

[YueHan0525]

yeah，i find it， At 2019-03-25 14:45:57, "Avisek Lahiri" <notifications@github.com> wrote: Hi. Could you solve this issue? I am facing a similar problem. I tried to train SRRESNET with MSE loss which looked fine but when I tried to initiate SRAGN phase with the vgg54 loss I also got a similar kind of error message. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[avisekiit]

modified 74 rows "return layers[layer]" in srgan.py

Hi. Just confirming. With this fix you

1. First trained SRRESNET with MSE loss and then
2. Trained SRAGN with vgg54 loss + adversarial loss ?

Did you do anything different than these 2 steps?

Thanks again for your reply.

[YueHan0525]

My work is that
First trained SRRESNET with VGG22 loss and then trained SRGAN with VGG54 loss,but worsen them for vgg22.

[avisekiit]

My work is that
First trained SRRESNET with VGG22 loss and then trained SRGAN with VGG54 loss,but worsen them for vgg22.

Hi. Thanks for the reply. Did you get good results by training first SRRESNET with VGG22 and then SRGAN with VGG54 + adversarial loss ?
Can you please let me know the command line statements that you used for running the codes. I also ran in this sequence but getting bad results. So, want to confirm that I am also running the codes like you did. Thanks in advance.

[YueHan0525]

I also got bad results.No reason found.