Unadversarial Examples: Designing Objects for Robust Vision
Hadi Salman111Equal contribution. Microsoft Research Andrew Ilyas MIT Logan Engstrom MIT Sai Vemprala Microsoft Research Aleksander Mądry MIT Ashish Kapoor Microsoft Research
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.
中文速览
现实中许多计算机视觉任务允许设计者同时控制模型和被识别的物体，研究者正是利用这一点，提出了"反对抗样本"（unadversarial examples）框架：不再像对抗攻击那样扰动输入来欺骗模型，而是反过来优化物体本身的贴纸（patch）或三维纹理（texture），让模型能更自信、更准确地识别它们。具体做法是借鉴对抗样本领域的投影梯度下降方法，最小化而非最大化分类器的损失函数，从而生成在各种随机位置、角度和图像损坏条件下仍能稳健发挥作用的视觉标记。实验覆盖了 CIFAR、ImageNet 等标准基准及其损坏版本、仿真无人机着陆场景和真实物理世界，结果显示这些优化过的贴纸或纹理能将分类准确率提升超过 26%，同时显著增强对未见损坏类型的鲁棒性。这项工作表明，当系统设计者拥有对物体外观的部分控制权时，可以主动"为机器设计物体"，从而将深度学习对输入敏感这一传统弱点转变为提升性能的新途径。
原文 arXiv:2012.12235;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2012.12235v1