Comment by egypturnash

How does this work? Have you tried training anything on images processed with this and verified that whatever you're doing actually glitches up the resulting datasets?