RaSP: Relation-aware Semantic Prior for Weakly Supervised Incremental Segmentation

TL;DR

This paper introduces RaSP, a weakly supervised method leveraging semantic relations to improve incremental segmentation with only image-level labels, reducing annotation costs and enhancing segmentation quality.

Contribution

It proposes a novel relation-aware semantic prior that transfers objectness information between classes, improving weakly supervised incremental segmentation performance.

Findings

Semantic relations improve segmentation accuracy.

Even simple class interactions enhance old and new class segmentation.

Method remains effective in longer, realistic task sequences.

Abstract

Class-incremental semantic image segmentation assumes multiple model updates, each enriching the model to segment new categories. This is typically carried out by providing expensive pixel-level annotations to the training algorithm for all new objects, limiting the adoption of such methods in practical applications. Approaches that solely require image-level labels offer an attractive alternative, yet, such coarse annotations lack precise information about the location and boundary of the new objects. In this paper we argue that, since classes represent not just indices but semantic entities, the conceptual relationships between them can provide valuable information that should be leveraged. We propose a weakly supervised approach that exploits such semantic relations to transfer objectness prior from the previously learned classes into the new ones, complementing the supervisory signal from image-level labels. We validate our approach on a number of continual learning tasks, and show how even a simple pairwise interaction between classes can significantly improve the segmentation mask quality of both old and new classes. We show these conclusions still hold for longer and, hence, more realistic sequences of tasks and for a challenging few-shot scenario.