Image Synthesis with a Single (Robust) Classifier

June 06, 2019 · Entered Twilight · 🏛 Neural Information Processing Systems

"No code URL or promise found in abstract"
"Code repo scraped from project page (backfill)"

Evidence collected by the PWNC Scanner

Repo contents: README.md, generation.ipynb, headline.jpg, image_to_image_translation.ipynb, inpainting.ipynb, paint_with_features.ipynb, requirements.txt, sample_inputs, sketch_to_image.ipynb, superresolution.ipynb, user_constants.py

Authors Shibani Santurkar, Dimitris Tsipras, Brandon Tran, Andrew Ilyas, Logan Engstrom, Aleksander Madry arXiv ID 1906.09453 Category cs.CV: Computer Vision Cross-listed cs.LG, cs.NE, stat.ML Citations 36 Venue Neural Information Processing Systems Repository https://github.com/MadryLab/robustness_applications.git ⭐ 129 Last Checked 13 days ago

Abstract

We show that the basic classification framework alone can be used to tackle some of the most challenging tasks in image synthesis. In contrast to other state-of-the-art approaches, the toolkit we develop is rather minimal: it uses a single, off-the-shelf classifier for all these tasks. The crux of our approach is that we train this classifier to be adversarially robust. It turns out that adversarial robustness is precisely what we need to directly manipulate salient features of the input. Overall, our findings demonstrate the utility of robustness in the broader machine learning context. Code and models for our experiments can be found at https://git.io/robust-apps.