Scalable Certified Segmentation via Randomized Smoothing
This work addresses the need for scalable and statistically sound certified segmentation methods in computer vision, offering a practical solution for safety-critical applications.
The paper tackles the problem of providing certified robustness guarantees for image and point cloud segmentation by introducing a new method based on randomized smoothing, achieving competitive accuracy and certification on real-world datasets like Pascal Context, Cityscapes, and ShapeNet.
We present a new certification method for image and point cloud segmentation based on randomized smoothing. The method leverages a novel scalable algorithm for prediction and certification that correctly accounts for multiple testing, necessary for ensuring statistical guarantees. The key to our approach is reliance on established multiple-testing correction mechanisms as well as the ability to abstain from classifying single pixels or points while still robustly segmenting the overall input. Our experimental evaluation on synthetic data and challenging datasets, such as Pascal Context, Cityscapes, and ShapeNet, shows that our algorithm can achieve, for the first time, competitive accuracy and certification guarantees on real-world segmentation tasks. We provide an implementation at https://github.com/eth-sri/segmentation-smoothing.