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 classiï¬er 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 classiï¬er to adapt to the target domain, we use a separable component called data calibrator to help the ï¬xed source classiï¬er recover discrimination power in the target domain, while preserving the source domain’s performance. When the difference between two domains is small, the source classiï¬er’s representation is sufï¬cient 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 also empirically suggests the potential connection between domain adaptation and adversarial attacks. Code release is available at https://github.com/yeshaokai/ Calibrator-Domain-Adaptation