Taxonomy-Aware Continual Semantic Segmentation in Hyperbolic Spaces for Open-World Perception

作     者：Hindel, Julia Cattaneo, Daniele Valada, Abhinav 

作者机构：Univ Freiburg Dept Comp Sci D-79110 Freiburg Germany 

出 版 物：《IEEE ROBOTICS AND AUTOMATION LETTERS》 (IEEE Robot. Autom.)

年 卷 期：2025年第10卷第2期

页      面：1904-1911页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0811[工学-控制科学与工程] 

基　　金：Deutsche Forschungsgemeinschaft (DFG  German Research Foundation) [SFB 1597 - 499552394] 

主　　题：Semantic segmentation Semantics Taxonomy Data models Incremental learning Training Rigidity Adaptation models Visualization Predictive models Continual learning incremental learning deep learning for visual perception 

摘      要：Semantic segmentation models are typically trained on a fixed set of classes, limiting their applicability in open-world scenarios. Class-incremental semantic segmentation aims to update models with emerging new classes while preventing catastrophic forgetting of previously learned ones. However, existing methods impose strict rigidity on old classes, reducing their effectiveness in learning new incremental classes. In this work, we propose Taxonomy-Oriented Poincar & eacute;-regularized Incremental-Class Segmentation (TOPICS) that learns feature embeddings in hyperbolic space following explicit taxonomy-tree structures. This supervision provides plasticity for old classes, updating ancestors based on new classes while integrating new classes at fitting positions. Additionally, we maintain implicit class relational constraints on the geometric basis of the Poincar & eacute;ball. This ensures that the latent space can continuously adapt to new constraints while maintaining a robust structure to combat catastrophic forgetting. We also establish eight realistic incremental learning protocols for autonomous driving scenarios, where novel classes can originate from known classes or the background. Extensive evaluations of TOPICS on the Cityscapes and Mapillary Vistas 2.0 benchmarks demonstrate that it achieves state-of-the-art performance.