Threshold-adaptive Unsupervised Focal Loss for Domain Adaptation of Semantic Segmentation
This addresses the domain shift problem for semantic segmentation in autonomous driving, enabling better performance without manual labeling, but it is incremental as it builds on existing UDA methods.
The paper tackles the problem of unsupervised domain adaptation for semantic segmentation, where models trained on simulated data perform poorly in real scenes, by proposing a two-stage entropy-based method with a threshold-adaptive unsupervised focal loss and cross-domain image mixing, achieving state-of-the-art mIoUs of 58.4% and 59.6% on SYNTHIA-to-Cityscapes and GTA5-to-Cityscapes using DeepLabV2.
Semantic segmentation is an important task for intelligent vehicles to understand the environment. Current deep learning methods require large amounts of labeled data for training. Manual annotation is expensive, while simulators can provide accurate annotations. However, the performance of the semantic segmentation model trained with the data of the simulator will significantly decrease when applied in the actual scene. Unsupervised domain adaptation (UDA) for semantic segmentation has recently gained increasing research attention, aiming to reduce the domain gap and improve the performance on the target domain. In this paper, we propose a novel two-stage entropy-based UDA method for semantic segmentation. In stage one, we design a threshold-adaptative unsupervised focal loss to regularize the prediction in the target domain, which has a mild gradient neutralization mechanism and mitigates the problem that hard samples are barely optimized in entropy-based methods. In stage two, we introduce a data augmentation method named cross-domain image mixing (CIM) to bridge the semantic knowledge from two domains. Our method achieves state-of-the-art 58.4% and 59.6% mIoUs on SYNTHIA-to-Cityscapes and GTA5-to-Cityscapes using DeepLabV2 and competitive performance using the lightweight BiSeNet.