Although current semantic segmentation approaches have achieved impressive performance, their ability to incrementally learn new classes is limited. Moreover, pixel-by-pixel annotations are costly and time-consuming. Therefore, a new field called Weakly-supervised Incremental Learning for Semantic Segmentation (WILSS) has emerged, which learns new classes using image-level labels. However, image-level labels do not provide sufficient detail, and we discover that the state-of-the-art of WILSS suffers from confusion between old knowledge and new knowledge. To address this issue, we propose eakly-supervised ncremental learning for emantic segmentation with Class ierarchy (WISH), a method that considers the hierarchical structure of each class when determining which knowledge to trust in cases of confusion between old and new knowledge. Our method has achieved new state-of-the-art performances in all settings compared to the previous methods on the Pascal VOC and MS COCO datasets.

中文翻译：

---

具有类层次结构的语义分割的弱监督增量学习

尽管当前的语义分割方法已经取得了令人印象深刻的性能，但它们增量学习新类的能力是有限的。此外，逐像素注释既昂贵又耗时。因此，出现了一个名为弱监督语义分割增量学习（WILSS）的新领域，它使用图像级标签来学习新类别。然而，图像级标签没有提供足够的细节，我们发现 WILSS 的最新技术存在旧知识和新知识之间的混淆。为了解决这个问题，我们提出了带有类层次结构（WISH）的语义分割的弱监督增量学习，这种方法在新旧知识混淆的情况下确定信任哪些知识时考虑每个类的层次结构。与 Pascal VOC 和 MS COCO 数据集上的先前方法相比，我们的方法在所有设置中都实现了新的最先进的性能。