SG-One: Similarity Guidance Network for One-Shot Semantic Segmentation

TL;DR

This paper introduces SG-One, a novel similarity guidance network for one-shot semantic segmentation that uses cosine similarity and a unified end-to-end framework to effectively segment unseen categories from a single support example.

Contribution

The paper proposes a simple, effective framework that employs masked average pooling and cosine similarity for robust one-shot segmentation, outperforming baseline methods on Pascal VOC 2012.

Findings

Achieves 46.3% mIoU on Pascal VOC 2012

Utilizes a unified end-to-end network for support and query images

Surpasses baseline methods in one-shot segmentation performance

Abstract

One-shot image semantic segmentation poses a challenging task of recognizing the object regions from unseen categories with only one annotated example as supervision. In this paper, we propose a simple yet effective Similarity Guidance network to tackle the One-shot (SG-One) segmentation problem. We aim at predicting the segmentation mask of a query image with the reference to one densely labeled support image of the same category. To obtain the robust representative feature of the support image, we firstly adopt a masked average pooling strategy for producing the guidance features by only taking the pixels belonging to the support image into account. We then leverage the cosine similarity to build the relationship between the guidance features and features of pixels from the query image. In this way, the possibilities embedded in the produced similarity maps can be adapted to guide the process of segmenting objects. Furthermore, our SG-One is a unified framework which can efficiently process both support and query images within one network and be learned in an end-to-end manner. We conduct extensive experiments on Pascal VOC 2012. In particular, our SGOne achieves the mIoU score of 46.3%, surpassing the baseline methods.