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

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 Even though a large number of methods propose new adaptation strategies, they are mostly based on outdated network architectures.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 multilevel 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 -> Cityscapes and 5.4 mIoU for Synthia -> Cityscapes and enables learning even difficult classes such as train, bus, and truck well. The implementation is available at https://github.com/lhoyer/DAFormer.

Automated summary

As acquiring pixel-wise annotations for real-world images is costly, this paper explores the use of synthetic data and unsupervised domain adaptation (UDA) to train a model that can adapt to real images without annotations. The authors benchmark different network architectures for UDA and find the potential of Transformers for UDA semantic segmentation. They propose a novel UDA method called DAFormer, which achieves significant improvements in state-of-the-art results.