Abstract:
Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classier 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 classier to adapt to the target domain, we use a separable component called data calibrator to help the xed source classier recover discrimination power in the target domain, while preserving the source domains performance. When the difference between two domains is small, the source classiers representation is sufcient 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
