program to remove glaze / nightshade / glaze2 from images
pip install -r requirements.txt
python run.py -h
python run.py clean image.png
python run.py clean_folder path/to/images --newfolder path/to/cleaned/images
- vaeloop
+ most consistent
~ passes image through VAE, which removes imperfections (e.g. glaze)
~ resizes images to a mult of 8
- needs a gpu- glaze1
~ classical glaze remover (has it even worked at all?)- glaze2
+ both vaeloop and glaze1 combined
~ passes image through VAE, which removes imperfections (e.g. glaze)
~ resizes images to a mult of 8
- needs a gpuconsidering that glaze/nightshade doesnt work at all, this is extremely effective (100% captioning accuracy after deglazing)
training effectiveness decreases after deglazing since glaze actually helps training (the glaze/shade acts as noise offset)
basically nothing; it adds adversarial noise which supposedly makes the model think a dog is a cat
the issue is SD doesnt really care about that since we pass it through a VAE, but it does confuse the CLIP (see: nightshade)
vaeloop will fix the 'glaze' and make it CLIP taggable (even though most people dont use CLIP)
glaze n co think that they've made a miracle cure for ai training on other art; they have not
not without making it terrible to watch/private
pick 2 of 3 things:
- untrainable
- good to watch
- publicly viewable