Uncertainty Estimation via Response Scaling for Pseudo-mask Noise Mitigation in Weakly-supervised Semantic Segmentation

TL;DR

This paper introduces URN, a novel method for estimating uncertainty in weakly-supervised semantic segmentation by response scaling, which effectively mitigates noise in pseudo-masks and improves segmentation accuracy.

Contribution

The paper proposes a new uncertainty estimation technique via response scaling (URN) to reduce noise impact in pseudo-mask based WSSS, achieving state-of-the-art results.

Findings

URN improves segmentation accuracy on PASCAL VOC 2012.

URN achieves 71.2% mIoU on PASCAL VOC 2012.

URN achieves 41.5% mIoU on MS COCO 2014.

Abstract

Weakly-Supervised Semantic Segmentation (WSSS) segments objects without a heavy burden of dense annotation. While as a price, generated pseudo-masks exist obvious noisy pixels, which result in sub-optimal segmentation models trained over these pseudo-masks. But rare studies notice or work on this problem, even these noisy pixels are inevitable after their improvements on pseudo-mask. So we try to improve WSSS in the aspect of noise mitigation. And we observe that many noisy pixels are of high confidence, especially when the response range is too wide or narrow, presenting an uncertain status. Thus, in this paper, we simulate noisy variations of response by scaling the prediction map multiple times for uncertainty estimation. The uncertainty is then used to weight the segmentation loss to mitigate noisy supervision signals. We call this method URN, abbreviated from Uncertainty estimation via Response scaling for Noise mitigation. Experiments validate the benefits of URN, and our method achieves state-of-the-art results at 71.2% and 41.5% on PASCAL VOC 2012 and MS COCO 2014 respectively, without extra models like saliency detection. Code is available at https://github.com/XMed-Lab/URN.