mks0601
commented
5 years ago The output of resnet is //32, //32. I use three 2x deconvolutional layer to upsample it by 8, which becomes //4, //4. For more information, please read my paper.
Closed ghost closed 4 years ago
mks0601
commented
5 years ago The output of resnet is //32, //32. I use three 2x deconvolutional layer to upsample it by 8, which becomes //4, //4. For more information, please read my paper.
Thank your implementation. I have a question about input shape. I looked into the code in main/config.py as below
The question is why out_shape is input_shape // 4 ?
If I remembered correctly, the input shape size ( width, height ) will be 8 times smaller (width/8, height/8) after feeding it to Resnet network. It confused me a lot time :(. Thank you.