SS-SFDA : Self-Supervised Source-Free Domain Adaptation for Road Segmentation in Hazardous Environments

TL;DR

This paper introduces SS-SFDA, a self-supervised source-free domain adaptation method for robust road segmentation in adverse weather, outperforming previous methods in accuracy and efficiency.

Contribution

The paper proposes a novel self-supervised SFDA algorithm with techniques like pseudo-labels, self-attention, curriculum learning, entropy minimization, and model distillation.

Findings

Outperforms prior SFDA methods by at least 10.26% in accuracy.

Reduces training time by 18-180 times.

Achieves accuracy comparable to supervised methods.

Abstract

We present a novel approach for unsupervised road segmentation in adverse weather conditions such as rain or fog. This includes a new algorithm for source-free domain adaptation (SFDA) using self-supervised learning. Moreover, our approach uses several techniques to address various challenges in SFDA and improve performance, including online generation of pseudo-labels and self-attention as well as use of curriculum learning, entropy minimization and model distillation. We have evaluated the performance on $6$ datasets corresponding to real and synthetic adverse weather conditions. Our method outperforms all prior works on unsupervised road segmentation and SFDA by at least 10.26%, and improves the training time by 18-180x. Moreover, our self-supervised algorithm exhibits similar accuracy performance in terms of mIOU score as compared to prior supervised methods.