Multi-similarity based hyperrelation network for few-shot segmentation

Abstract

Few-shot semantic segmentation aims at recognizing the object regions of unseen categories with only a few annotated examples as supervision. The key to few-shot segmentation is to establish a robust semantic relationship between the support and query images and to prevent overfitting. In this paper, an effective multi-similarity hyperrelation network (MSHNet) is proposed to tackle the few-shot semantic segmentation problem. In MSHNet, a new generative prototype similarity (GPS) is proposed, which, together with cosine similarity, establishes a strong semantic relationship between supported images and query images. In addition, a symmetric merging block (SMB) in MSHNet is proposed to efficiently merge multi-layer, multi-shot, multi-similarity features to generate hyperrelation features for semantic segmentation. Experimenting on two benchmark semantic segmentation datasets (Pascal − 5i and COCO − 20i) shows that this method achieves a mean intersection-over-union score of 72.3% and 56.0%, respectively, which outperforms the state-of-the-art methods by 1.9% and 6.5%.