Tweeted By @hardmaru
Computer Vision with a Single Robust Classifier
— hardmaru (@hardmaru) June 11, 2019
Interesting work: “Using 𝑜𝑛𝑙𝑦 an adversarially trained classifier (standard ResNet w/ cross-entropy), we show that we can perform image generation, super resolution, inpainting, and interactive editing”https://t.co/VtmUohz2sZ https://t.co/6c1yxeZgJ9