DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation

TL;DR

DAFormer introduces a Transformer-based architecture with novel training strategies for unsupervised domain adaptation in semantic segmentation, significantly improving performance over previous methods and effectively handling difficult classes.

Contribution

It systematically benchmarks network architectures for UDA and proposes DAFormer, a Transformer-based model with training strategies that enhance domain adaptation performance.

Findings

DAFormer outperforms previous methods by 10.8 mIoU on GTA-to-Cityscapes.

It achieves a 5.4 mIoU improvement on Synthia-to-Cityscapes.

The model effectively learns difficult classes like train, bus, and truck.

Abstract

As acquiring pixel-wise annotations of real-world images for semantic segmentation is a costly process, a model can instead be trained with more accessible synthetic data and adapted to real images without requiring their annotations. This process is studied in unsupervised domain adaptation (UDA). Even though a large number of methods propose new adaptation strategies, they are mostly based on outdated network architectures. As the influence of recent network architectures has not been systematically studied, we first benchmark different network architectures for UDA and newly reveal the potential of Transformers for UDA semantic segmentation. Based on the findings, we propose a novel UDA method, DAFormer. The network architecture of DAFormer consists of a Transformer encoder and a multi-level context-aware feature fusion decoder. It is enabled by three simple but crucial training strategies to stabilize the training and to avoid overfitting to the source domain: While (1) Rare Class Sampling on the source domain improves the quality of the pseudo-labels by mitigating the confirmation bias of self-training toward common classes, (2) a Thing-Class ImageNet Feature Distance and (3) a learning rate warmup promote feature transfer from ImageNet pretraining. DAFormer represents a major advance in UDA. It improves the state of the art by 10.8 mIoU for GTA-to-Cityscapes and 5.4 mIoU for Synthia-to-Cityscapes and enables learning even difficult classes such as train, bus, and truck well. The implementation is available at https://github.com/lhoyer/DAFormer.