Re-distributing Biased Pseudo Labels for Semi-supervised Semantic Segmentation: A Baseline Investigation

TL;DR

This paper introduces DARS, a simple method for semi-supervised semantic segmentation that reduces class bias in pseudo labels through distribution alignment and sampling, improving performance on standard datasets.

Contribution

The paper proposes a novel Distribution Alignment and Random Sampling (DARS) technique and a progressive augmentation strategy to produce unbiased pseudo labels in semi-supervised segmentation.

Findings

DARS effectively reduces class bias in pseudo labels.

The method outperforms state-of-the-art approaches on Cityscapes and PASCAL VOC 2012.

Code implementation is publicly available.

Abstract

While self-training has advanced semi-supervised semantic segmentation, it severely suffers from the long-tailed class distribution on real-world semantic segmentation datasets that make the pseudo-labeled data bias toward majority classes. In this paper, we present a simple and yet effective Distribution Alignment and Random Sampling (DARS) method to produce unbiased pseudo labels that match the true class distribution estimated from the labeled data. Besides, we also contribute a progressive data augmentation and labeling strategy to facilitate model training with pseudo-labeled data. Experiments on both Cityscapes and PASCAL VOC 2012 datasets demonstrate the effectiveness of our approach. Albeit simple, our method performs favorably in comparison with state-of-the-art approaches. Code will be available at https://github.com/CVMI-Lab/DARS.