Towards Better Certified Segmentation via Diffusion Models - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2023

Towards Better Certified Segmentation via Diffusion Models

Abstract

The robustness of image segmentation has been an important research topic in the past few years as segmentation models have reached productionlevel accuracy. However, like classification models, segmentation models can be vulnerable to adversarial perturbations, which hinders their use in criticaldecision systems like healthcare or autonomous driving. Recently, randomized smoothing has been proposed to certify segmentation predictions by adding Gaussian noise to the input to obtain theoretical guarantees. However, this method exhibits a trade-off between the amount of added noise and the level of certification achieved. In this paper, we address the problem of certifying segmentation prediction using a combination of randomized smoothing and diffusion models. Our experiments show that combining randomized smoothing and diffusion models significantly improves certified robustness, with results indicating a mean improvement of 21 points in accuracy compared to previous state-of-the-art methods on Pascal-Context and Cityscapes public datasets. Our method is independent of the selected segmentation model and does not need any additional specialized training procedure.
Fichier principal
Vignette du fichier
laousy_507.pdf (15.52 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04388219 , version 1 (11-01-2024)

Identifiers

Cite

Othmane Laousy, Alexandre Araujo, Guillaume Chassagnon, Marie-Pierre Revel, Siddharth Garg, et al.. Towards Better Certified Segmentation via Diffusion Models. UAI 2023 - The 39th Conference on Uncertainty in Artificial Intelligence -, Jul 2023, Pittsburgh (Pennsylvania), United States. pp.1185-1195. ⟨hal-04388219⟩
40 View
4 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More