Unadversarial Examples: Designing Objects for Robust Vision
arXiv:2012.12235
Abstract
We study a class of realistic computer vision settings wherein one can influence the design of the objects being recognized. We develop a framework that leverages this capability to significantly improve vision models' performance and robustness. This framework exploits the sensitivity of modern machine learning algorithms to input perturbations in order to design "robust objects," i.e., objects that are explicitly optimized to be confidently detected or classified. We demonstrate the efficacy of the framework on a wide variety of vision-based tasks ranging from standard benchmarks, to (in-simulation) robotics, to real-world experiments. Our code can be found at https://git.io/unadversarial .
References in corpus (11)
- ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
- Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
- ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
- Exploring the Landscape of Spatial Robustness
- DARTS: Deceiving Autonomous Cars with Toxic Signs
- Adversarial Training Can Hurt Generalization
- On Physical Adversarial Patches for Object Detection
- Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations
- Physical Adversarial Examples for Object Detectors
- Testing Robustness Against Unforeseen Adversaries
- With Friends Like These, Who Needs Adversaries?