Scalable Certified Segmentation via Randomized Smoothing
Marc Fischer 1 Maximilian Baader 1 Martin Vechev 1
Abstract
We present a new certification method for image
and point cloud segmentation based on random-
ized smoothing. The method leverages a novel
scalable algorithm for prediction and certification (a) Attacked image (b) Ground truth segmentation
that correctly accounts for multiple testing, nec-
essary for ensuring stat ...


雷达卡




京公网安备 11010802022788号







