Out-of-Distribution Segmentation via Wasserstein-Based Evidential Uncertainty

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Brosch, Arnold, Eldesokey, Abdelrahman, Felsberg, Michael, Maag, Kira
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912759844700160
author Brosch, Arnold
Eldesokey, Abdelrahman
Felsberg, Michael
Maag, Kira
author_facet Brosch, Arnold
Eldesokey, Abdelrahman
Felsberg, Michael
Maag, Kira
contents Deep neural networks achieve superior performance in semantic segmentation, but are limited to a predefined set of classes, which leads to failures when they encounter unknown objects in open-world scenarios. Recognizing and segmenting these out-of-distribution (OOD) objects is crucial for safety-critical applications such as automated driving. In this work, we present an evidence segmentation framework using a Wasserstein loss, which captures distributional distances while respecting the probability simplex geometry. Combined with Kullback-Leibler regularization and Dice structural consistency terms, our approach leads to improved OOD segmentation performance compared to uncertainty-based approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2512_11373
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Out-of-Distribution Segmentation via Wasserstein-Based Evidential Uncertainty
Brosch, Arnold
Eldesokey, Abdelrahman
Felsberg, Michael
Maag, Kira
Computer Vision and Pattern Recognition
Machine Learning
Deep neural networks achieve superior performance in semantic segmentation, but are limited to a predefined set of classes, which leads to failures when they encounter unknown objects in open-world scenarios. Recognizing and segmenting these out-of-distribution (OOD) objects is crucial for safety-critical applications such as automated driving. In this work, we present an evidence segmentation framework using a Wasserstein loss, which captures distributional distances while respecting the probability simplex geometry. Combined with Kullback-Leibler regularization and Dice structural consistency terms, our approach leads to improved OOD segmentation performance compared to uncertainty-based approaches.
title Out-of-Distribution Segmentation via Wasserstein-Based Evidential Uncertainty
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2512.11373