Reproduce “ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Shang-Tse Chen, Cory Cornelius, Jason Martin, Duen Horng Chau

ShapeShifter adapts an existing robustness technique to attack Faster R-CNN object detection by optimizing a printable stop-sign alteration across simulated views and candidate boxes. In a parking-lot drive-by, the tested model labeled the altered sign “person” in 190 of 405 frames, but correctly labeled it “stop sign” once. This was a white-box test, not a deployed vehicle.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.