Structured Consistency Loss for semi-supervised semantic segmentation

TL;DR

This paper introduces a structured consistency loss for semi-supervised semantic segmentation, improving inter-pixel similarity consistency between teacher and student networks, and achieves state-of-the-art results on Cityscapes.

Contribution

It proposes a novel structured consistency loss tailored for semantic segmentation, addressing limitations of pixel-wise approaches and enhancing semi-supervised learning performance.

Findings

Achieved 81.9 mIoU on Cityscapes validation set.

Achieved 83.84 mIoU on Cityscapes test set.

Ranked first on Cityscapes semantic labeling benchmark.

Abstract

The consistency loss has played a key role in solving problems in recent studies on semi-supervised learning. Yet extant studies with the consistency loss are limited to its application to classification tasks; extant studies on semi-supervised semantic segmentation rely on pixel-wise classification, which does not reflect the structured nature of characteristics in prediction. We propose a structured consistency loss to address this limitation of extant studies. Structured consistency loss promotes consistency in inter-pixel similarity between teacher and student networks. Specifically, collaboration with CutMix optimizes the efficient performance of semi-supervised semantic segmentation with structured consistency loss by reducing computational burden dramatically. The superiority of proposed method is verified with the Cityscapes; The Cityscapes benchmark results with validation and with test data are 81.9 mIoU and 83.84 mIoU respectively. This ranks the first place on the pixel-level semantic labeling task of Cityscapes benchmark suite. To the best of our knowledge, we are the first to present the superiority of state-of-the-art semi-supervised learning in semantic segmentation.