Reproduce “Lipschitz Optimization for Formal Verification of Homographies” 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
Jean-Guillaume Durand, Panagiotis Kouvaros, Maxime Gariel, Alessio Lomuscio
The authors extend verification to camera-induced perspective changes by deriving homographies for planar scenes and bounding pixel values with piecewise-linear bounds refined by Lipschitz optimization. The method assumes a pinhole camera, known lens distortion, and negligible parallax and occlusion. On a 2-layer MNIST VNN-COMP model, 25% of test images were certified robust to yaw over [0°, 5°] under the stated setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.