search

soapisnotfat / super-resolution

collection of super-resolution models & algorithms
Apache License 2.0
744 stars 188 forks source link

confirm SRCNN #9

Closed micklexqg closed 5 years ago

micklexqg commented 5 years ago

Hi, I am running your code. but I am a little puzzled about the SRCNN. generally, the size of input is same as that of output as shown: image but i do not find upsampling the resized or cropped batch. does the srcnn lack pre-processing?

soapisnotfat commented 5 years ago

Hi, this graph is misleading but the code should be correct.

soapisnotfat commented 5 years ago

fixed the second 1x1 filters in SRCNN. Thanks!

micklexqg commented 5 years ago

thanks for reply, I think that it is right with the srcnn model, but lack two resize operation to produce training data and labels. it is just ok to add this preprocess.

soapisnotfat commented 5 years ago

Yeah, I intentionally delete the resize operations from the original implementation for fitting my unified dataset.