search

lltcggie / waifu2x-caffe

waifu2xのCaffe版
MIT License
8.09k stars 839 forks source link

waifu2x-caffe-cui fail to learn image model #176

Open dummy186 opened 5 years ago

dummy186 commented 5 years ago

I entered this command in waifu2x-caffe-cui. waifu2x-caffe-cui -model_dir models/my_model -m noise -noise_level 1 -i learn/test001.jpg -o output.png

But, it is not work, and print error message 変換に失敗したファイルがあります. All file is exists. What is problem? Is it a feature not originally supported?

nagadomi commented 5 years ago

waifu2x-caffe does not support training. If your model(models/my_model) is trained with my lua implementation, you can use it with info.json, noise1_model.prototxt and noise1_model.json properly configured.

DJDiX35 commented 5 years ago

Can you add learning to caffe, or instruction how to do it? I'm good with photoshop and after increasing images to 7-8x assembling one good picture from mixing of 4 pictures from every red.noise mode of your program (cunet x7-x8 + every of 4 noise reduction levels) An wanna to use this "merged" in photoshop pictures to increase other results. It's not only for me - it's for all, because if results will be better - program will have more popularity. (Using windows-assembled waifu2x-caffe)