Out-of-Distribution Segmentation via Wasserstein-Based Evidential Uncertainty
This addresses safety-critical applications like automated driving by enhancing segmentation for unknown objects, but it appears incremental as it builds on existing uncertainty-based approaches.
The paper tackles the problem of segmenting unknown objects in open-world scenarios by proposing an evidence segmentation framework with a Wasserstein loss, achieving improved out-of-distribution segmentation performance compared to uncertainty-based methods.
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.