When I ran the test code, I realized that the reconstructed image wasn't quite right

huzi96 / Coarse2Fine-PyTorch

70 stars 6 forks source link

When I ran the test code, I realized that the reconstructed image wasn't quite right #3

Open CaiShilv opened 3 years ago

CaiShilv commented 3 years ago

Wouldn't the network function torch. Round () change all the values to 0 when running forward?When I ran the test code, I realized that the reconstructed image wasn't quite right.

huzi96 commented 3 years ago

The round() function does not change values to 0. Instead, it keeps the integer part of a real value (e.g. 1.234 -> 1.000). May I see an example of what went wrong with the reconstructed image? Are you running the encoder and the decoder on different hardware?

CaiShilv commented 3 years ago

I have problems with my own training weights. Can you provide some weights for the PyTorch version?

CaiShilv commented 3 years ago

I'm so sorry,I discovered that there was a problem when I was loading the model.Thanks you very much!

huzi96 commented 3 years ago

Glad to hear that you solved the problem ;)