search

jmiller656 / EDSR-Tensorflow

Tensorflow implementation of Enhanced Deep Residual Networks for Single Image Super-Resolution
MIT License
330 stars 107 forks source link

the problem in using my own medical images for training #19

Closed chris1992212 closed 5 years ago

chris1992212 commented 6 years ago

hey, I have read and used your code for my own data. And my data is not easy to converge, So here is some questions about your code. First, why do you remove the last convention after the up sampling? Second, as I know, we often normalize our data in the range of 0 to 1, but in your code, the range is from 0 to 255. Thanks !

jmiller656 commented 5 years ago

The last convolution seemed to cause problems for me empirically, so I ended up removing it. I believe I chose my normalization procedure for images based on what they did in the EDSR paper