Computer Vision with a Single (Robust) Classifier
By Shibani Santurkar, Dimitris Tsipras, Brandon Tran, Andrew Ilyas, Logan Engstrom, Aleksander Madry. 2019
36 citations recorded by the source corpus.
We show that the basic classification framework alone can be used to tackle some of the most challenging computer vision tasks. 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.
Subfields: Artificial Intelligence, Biophysics, Computer Vision and Pattern Recognition
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Quoc V. Le
- Mohit Bansal
- Sergey Levine
- Nicholas Carlini
- Florian Tramèr
- Percy Liang
- Yoshua Bengio
- Pieter Abbeel
In Biophysics
- Kuan-Chieh Jackson Wang
- Aleksander Madry
- Mohammad Norouzi
- Yu Cheng
- Ting-Li Chen
- David Kristjanson Duvenaud
- Kevin Swersky
- Nicholas Carlini
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Ming-Hsuan Yang
- Bernt Schiele
- Jitendra Malik
- Xiaogang Wang
- Alan L. Yuille
Citations in the corpus
- Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One (2019, methodology): “With respect to theL2 norm, JEM performs competitively with adversarial training (Santurkar et al., 2019) and above the state-of-the-art certified robutness method of Salman et al. (2019) (“RandAdvSmooth” in Figure 5).”
- Adversarial Robustness: From Self-Supervised Pre-Training to Fine-Tuning (2020, methodology): “We follow [29] to choose 10 super classes which contain a total of 190 ImageNet classes.”
- The Origins and Prevalence of Texture Bias in Convolutional Neural Networks (2019, background): “State-of-the-art defenses on ImageNet use adversarial training [87, 41, 66, 100] or randomized smoothing [17], and images generated by optimizing class confidence under these models are more perceptually recognizable to humans [80, 30, 93, 51].”
- On Adaptive Attacks to Adversarial Example Defenses (2020, result): “That is, an unbounded attack that maximizes the score of a detector results in semantically meaningful images from that class—mirroring similar results for adversarially trained classifiers (Santurkar et al., 2019).”