Towards Realistic Incremental Scenario in Class Incremental Semantic Segmentation

arXiv:2405.09858v1 Announce Type: cross
Abstract: This paper addresses the unrealistic aspect of the commonly adopted Continuous Incremental Semantic Segmentation (CISS) scenario, termed overlapped. We point out that overlapped allows the same image to reappear in future tasks with different pixel labels, which is far from practical incremental learning scenarios. Moreover, we identified that this flawed scenario may lead to biased results for two commonly used techniques in CISS, pseudo-labeling and exemplar memory, resulting in unintended advantages or disadvantages for certain …

abstract arxiv class continuous cs.cv cs.lg future image incremental incremental learning labels paper pixel practical segmentation semantic tasks type