CamoEnv: Transferable and environment-consistent adversarial camouflage in autonomous driving

Open Webpage

Zijian Zhu, Xiao Yang, Hang Su, Shibao Zheng

Published: 01 Jan 2025, Last Modified: 12 Nov 2025Pattern Recognit. Lett. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Proposed CamoEnv blends seamlessly with varying viewpoints in dynamic environments.•Reduces overfitting and enhances transferability to black-box detectors.•Outperforms existing methods in transfer attack success rate.•Strong zero-shot transferability to tasks like segmentation and image grounding.