Frequency-based Matcher for Long-tailed Semantic Segmentation

TL;DR

This paper introduces a new benchmark and a transformer-based frequency matcher to address the long-tailed distribution problem in semantic segmentation, enhancing performance on underrepresented classes.

Contribution

It establishes the first comprehensive datasets and evaluation system for long-tailed semantic segmentation and proposes a novel frequency-based matcher to improve class matching.

Findings

The benchmark reveals significant performance gaps in existing methods.

The frequency-based matcher effectively reduces oversuppression of rare classes.

Experimental results demonstrate improved segmentation accuracy on long-tailed datasets.

Abstract

The successful application of semantic segmentation technology in the real world has been among the most exciting achievements in the computer vision community over the past decade. Although the long-tailed phenomenon has been investigated in many fields, e.g., classification and object detection, it has not received enough attention in semantic segmentation and has become a non-negligible obstacle to applying semantic segmentation technology in autonomous driving and virtual reality. Therefore, in this work, we focus on a relatively under-explored task setting, long-tailed semantic segmentation (LTSS). We first establish three representative datasets from different aspects, i.e., scene, object, and human. We further propose a dual-metric evaluation system and construct the LTSS benchmark to demonstrate the performance of semantic segmentation methods and long-tailed solutions. We also propose a transformer-based algorithm to improve LTSS, frequency-based matcher, which solves the oversuppression problem by one-to-many matching and automatically determines the number of matching queries for each class. Given the comprehensiveness of this work and the importance of the issues revealed, this work aims to promote the empirical study of semantic segmentation tasks. Our datasets, codes, and models will be publicly available.