Small Objects Matters in Weakly-supervised Semantic Segmentation

TL;DR

This paper introduces a new evaluation metric and dataset to assess weakly-supervised semantic segmentation methods across object sizes, revealing challenges with small objects and proposing a size-balanced loss to improve performance.

Contribution

The paper presents a novel size-balanced evaluation metric, a size-balanced dataset, and a loss function that enhances WSSS methods for small object segmentation.

Findings

Existing WSSS methods struggle with small objects.

The size-balanced loss improves segmentation accuracy across datasets.

Evaluation metrics reveal size-related performance gaps.

Abstract

Weakly-supervised semantic segmentation (WSSS) performs pixel-wise classification given only image-level labels for training. Despite the difficulty of this task, the research community has achieved promising results over the last five years. Still, current WSSS literature misses the detailed sense of how well the methods perform on different sizes of objects. Thus we propose a novel evaluation metric to provide a comprehensive assessment across different object sizes and collect a size-balanced evaluation set to complement PASCAL VOC. With these two gadgets, we reveal that the existing WSSS methods struggle in capturing small objects. Furthermore, we propose a size-balanced cross-entropy loss coupled with a proper training strategy. It generally improves existing WSSS methods as validated upon ten baselines on three different datasets.