BiMaL: Bijective Maximum Likelihood Approach to Domain Adaptation in Semantic Scene Segmentation

TL;DR

This paper introduces BiMaL, a novel bijective maximum likelihood loss for unsupervised domain adaptation in semantic segmentation, improving generalization across different datasets without assuming pixel independence.

Contribution

The paper proposes a new BiMaL loss and an Un-aligned Domain Score for better domain adaptation in semantic segmentation, outperforming state-of-the-art methods.

Findings

BiMaL outperforms SOTA on multiple domain adaptation benchmarks.

The Un-aligned Domain Score effectively measures model transferability.

BiMaL does not assume pixel independence, unlike previous methods.

Abstract

Semantic segmentation aims to predict pixel-level labels. It has become a popular task in various computer vision applications. While fully supervised segmentation methods have achieved high accuracy on large-scale vision datasets, they are unable to generalize on a new test environment or a new domain well. In this work, we first introduce a new Un-aligned Domain Score to measure the efficiency of a learned model on a new target domain in unsupervised manner. Then, we present the new Bijective Maximum Likelihood(BiMaL) loss that is a generalized form of the Adversarial Entropy Minimization without any assumption about pixel independence. We have evaluated the proposed BiMaL on two domains. The proposed BiMaL approach consistently outperforms the SOTA methods on empirical experiments on "SYNTHIA to Cityscapes", "GTA5 to Cityscapes", and "SYNTHIA to Vistas".