DALocNet: Improving Localization Accuracy for Domain Adaptive Object Detection

Abstract

Object detection assumes that the training data is identical to the testing data. However, the distributions of training and testing data, in practice, are different, thereby limiting the detection accuracy of objects. To solve this problem, recent works adopt domain adaptation techniques to reduce the domain discrepancy. In this paper, we present a novel deep neural network design for domain adaptive object detection by further improving the localization accuracy of objects. First, we present to refine the pseudo labels generated from the current object detection methods and use these labels with a weighted loss function to train the network on the target domain. Second, we insert several residual blocks into the shallow layers of a convolutional neural network used in the target domain to enhance detailed spatial information, which helps for object localization. We perform various experiments to evaluate our network on three widely-used public benchmark datasets for domain adaptive object detection. The experimental results show that our DALocNet performs favorably against the state-of-the-art methods on all the datasets quantitatively and qualitatively.