PADetBench: Towards benchmarking texture- and patch-based physical attacks against object detection

doi:10.1016/j.knosys.2025.114395

PADetBench: Towards benchmarking texture- and patch-based physical attacks against object detection

2025-09-04

0

PRE

AI

DOI：10.1016/j.knosys.2025.114395

Original

Original request for help

Save

Abstract

En 中文

• Comprehensive Benchmark Framework: We present the first large-scale benchmark integrating 23 physical attack methods and 48 object detection models, enabling systematic comparative analysis. • Realistic Simulation Environment: Our framework accurately models physical dynamics and cross-domain transformations, balancing reproducibility with practical relevance. • Extensive Empirical Analysis: Over 8,000 evaluations reveal critical insights into detector vulnerabilities and attack limitations. • Open-Source Infrastructure: Our end-to-end pipeline, with publicly available code and datasets, provides a foundation for future research in AI security.

PADetBench: Towards benchmarking texture- and patch-based physical attacks against object detection

Abstract

Journal

Organization