Interclass Prototype Relation for Few-Shot Segmentation
This work addresses the challenge of segmenting new classes with minimal annotations for computer vision applications, representing an incremental improvement over existing few-shot segmentation methods.
The paper tackles the problem of few-shot segmentation where sparse target class data makes classification boundaries difficult to set, especially for similar classes, and proposes IPRNet to improve separation by reducing interclass similarity, achieving state-of-the-art performance on Pascal-5i and COCO-20i benchmarks.
Traditional semantic segmentation requires a large labeled image dataset and can only be predicted within predefined classes. To solve this problem, few-shot segmentation, which requires only a handful of annotations for the new target class, is important. However, with few-shot segmentation, the target class data distribution in the feature space is sparse and has low coverage because of the slight variations in the sample data. Setting the classification boundary that properly separates the target class from other classes is an impossible task. In particular, it is difficult to classify classes that are similar to the target class near the boundary. This study proposes the Interclass Prototype Relation Network (IPRNet), which improves the separation performance by reducing the similarity between other classes. We conducted extensive experiments with Pascal-5i and COCO-20i and showed that IPRNet provides the best segmentation performance compared with previous research.