Certified Training for Convolutional Perturbations

Forum topic · 小凯 · 2026-07-22

Summary

This paper by Benedikt Brückner and Alessio Lomuscio (arXiv:2607.18195, cs.CV/cs.LG) introduces a novel certified training approach for making vision models provably robust to convolutional perturbations such as motion blur caused by camera shake. While data augmentation and adversarial training can improve empirical robustness, they lack formal safety guarantees, making hidden vulnerabilities hard to identify and mitigate. The proposed method leverages an efficient encoding of convolutional perturbations to train models with certifiable robustness guarantees. Experiments show the approach substantially outperforms adversarial training: on CIFAR10 it achieves over 80% robust accuracy against reasonable-strength motion blur while maintaining comparable standard accuracy. This addresses a key barrier to deploying vision models in safety-critical applications, where subtle perturbations like slight blurring can cause failures such as object detectors missing objects. The paper was auto-collected from zhichai.net on 2026-07-22.

Paper Overview

Research area: Computer Vision (CV) Authors: Benedikt Brückner, Alessio Lomuscio Published: 2026-07-20 arXiv: 2607.18195 Categories: cs.CV, cs.LG

Abstract

Vision models have been shown to be vulnerable to perturbations such as motion blur caused by camera shake. This hinders their deployment in critical applications, since phenomena like slight visual blurring can lead to failures — for example, object detectors missing objects. While methods such as data augmentation or adversarial training can improve empirical robustness, they lack formal safety guarantees, making it difficult to identify and mitigate hidden vulnerabilities.

The authors introduce a novel certified training approach that leverages an efficient encoding of convolutional perturbations to train provably robust models. Their method significantly outperforms adversarial training: for example, it achieves over 80% robust accuracy against reasonable-strength motion blur on CIFAR10, while maintaining comparable standard accuracy.

Refined Original Abstract

We introduce a Certified Training approach leveraging efficient encoding of convolutional perturbations, achieving over 80% robust accuracy against motion blur on CIFAR10 while maintaining comparable standard accuracy.