ESL: Entropy-guided Self-supervised Learning for Domain Adaptation in Semantic Segmentation

TL;DR

This paper introduces ESL, a novel entropy-guided self-supervised learning method that improves pseudo-label accuracy for unsupervised domain adaptation in semantic segmentation, leading to state-of-the-art results across benchmarks.

Contribution

ESL uses entropy as a confidence measure for pseudo-labeling, enhancing UDA performance in semantic segmentation over existing softmax-based methods.

Findings

ESL outperforms strong SSL baselines on multiple benchmarks.

ESL achieves state-of-the-art results in UDA for semantic segmentation.

Entropy-based confidence improves pseudo-label quality.

Abstract

While fully-supervised deep learning yields good models for urban scene semantic segmentation, these models struggle to generalize to new environments with different lighting or weather conditions for instance. In addition, producing the extensive pixel-level annotations that the task requires comes at a great cost. Unsupervised domain adaptation (UDA) is one approach that tries to address these issues in order to make such systems more scalable. In particular, self-supervised learning (SSL) has recently become an effective strategy for UDA in semantic segmentation. At the core of such methods lies `pseudo-labeling', that is, the practice of assigning high-confident class predictions as pseudo-labels, subsequently used as true labels, for target data. To collect pseudo-labels, previous works often rely on the highest softmax score, which we here argue as an unfavorable confidence measurement. In this work, we propose Entropy-guided Self-supervised Learning (ESL), leveraging entropy as the confidence indicator for producing more accurate pseudo-labels. On different UDA benchmarks, ESL consistently outperforms strong SSL baselines and achieves state-of-the-art results.