Not All Pixels Are Equal: Learning Pixel Hardness for Semantic Segmentation

TL;DR

This paper introduces a novel approach to learn pixel hardness in semantic segmentation by leveraging global and historical loss data, leading to improved performance and stability across various methods.

Contribution

It proposes a simple, effective method to learn pixel hardness maps that enhance segmentation accuracy without significant additional computational cost.

Findings

Achieves 1.37% mIoU improvement on Cityscapes

Demonstrates good cross-domain generalization

Applicable to most segmentation methods with minimal extra cost

Abstract

Semantic segmentation has recently witnessed great progress. Despite the impressive overall results, the segmentation performance in some hard areas (e.g., small objects or thin parts) is still not promising. A straightforward solution is hard sample mining, which is widely used in object detection. Yet, most existing hard pixel mining strategies for semantic segmentation often rely on pixel's loss value, which tends to decrease during training. Intuitively, the pixel hardness for segmentation mainly depends on image structure and is expected to be stable. In this paper, we propose to learn pixel hardness for semantic segmentation, leveraging hardness information contained in global and historical loss values. More precisely, we add a gradient-independent branch for learning a hardness level (HL) map by maximizing hardness-weighted segmentation loss, which is minimized for the segmentation head. This encourages large hardness values in difficult areas, leading to appropriate and stable HL map. Despite its simplicity, the proposed method can be applied to most segmentation methods with no and marginal extra cost during inference and training, respectively. Without bells and whistles, the proposed method achieves consistent/significant improvement (1.37% mIoU on average) over most popular semantic segmentation methods on Cityscapes dataset, and demonstrates good generalization ability across domains. The source codes are available at https://github.com/Menoly-xin/Hardness-Level-Learning .