Label Calibration for Semantic Segmentation Under Domain Shift

TL;DR

This paper introduces a fast, resource-efficient label calibration method for adapting pre-trained semantic segmentation models to new, unlabeled domains, significantly improving performance in synthetic-to-real transfer tasks.

Contribution

It proposes a novel prototype-based label calibration technique that adapts models to unlabelled target domains with minimal computational overhead.

Findings

Significant performance gains on synthetic-to-real segmentation tasks

Method is computationally efficient and easy to implement

Effective in scenarios with domain shift and unlabeled data

Abstract

Performance of a pre-trained semantic segmentation model is likely to substantially decrease on data from a new domain. We show a pre-trained model can be adapted to unlabelled target domain data by calculating soft-label prototypes under the domain shift and making predictions according to the prototype closest to the vector with predicted class probabilities. The proposed adaptation procedure is fast, comes almost for free in terms of computational resources and leads to considerable performance improvements. We demonstrate the benefits of such label calibration on the highly-practical synthetic-to-real semantic segmentation problem.