loss function

[githubjoiner]

why the loss function in your code is different from your paper when training the network.

[love6tao]

@githubjoiner the loss fuction in the code is binary_crossentropy, not ssim loss. the result images shown in the code look good. so this code just uses the encoder-decoder network in the paper

[love6tao]

the code of AE has an error. "encoded = Conv2D(1, (8, 8), strides=1, padding='same')(x)" the filter number shoud be 100, not 1

[nuclearboy95]

this is not an SSIM-AE,,,

[mozhuqing]

this is not an SSIM-AE,,,
Have you reproduced it? I've been watching this lately. Please.

[mozhuqing]

the code of AE has an error. "encoded = Conv2D(1, (8, 8), strides=1, padding='same')(x)" the filter number shoud be 100, not 1
Have you reproduced it? I've been watching this lately. Please.

[mozhuqing]

@githubjoiner the loss fuction in the code is binary_crossentropy, not ssim loss. the result images shown in the code look good. so this code just uses the encoder-decoder network in the paper

Have you reproduced it? I've been watching this lately. Please.