Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation

Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead of training a classifier to adapt to the target domain, we use a separable component called data calibrator to help the fixed source classifier recover discrimination power in the target domain, while preserving the source domain's performance. When the difference between two domains is small, the source classifier's representation is sufficient to perform well in the target domain and outperforms GAN-based methods in digits. Otherwise, the proposed method can leverage synthetic images generated by GANs to boost performance and achieve state-of-the-art performance in digits datasets and driving scene semantic segmentation. Our method empirically reveals that certain intriguing hints, which can be mitigated by adversarial attack to domain discriminators, are one of the sources for performance degradation under the domain shift.

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract

Results from the Paper

Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Synthetic-to-Real Translation GTAV-to-Cityscapes Labels Light-weight Calibrator mIoU 40.5 # 63
Domain Adaptation MNIST-to-USPS CyCleGAN (Light-weight Calibrator) Accuracy 97.1 # 7
Domain Adaptation SVHN-to-MNIST CyCleGAN (Light-weight Calibrator) Accuracy 97.5 # 5
Domain Adaptation USPS-to-MNIST CyCleGAN (Light-weight Calibrator) Accuracy 98.3 # 3


No methods listed for this paper. Add relevant methods here