Improving the Generalization Ability of Deep Neural Networks for Cross-Domain Visual Recognition
Jianwei Zheng, Chao Lü, Cong Hao, Deming Chen, Donghui Guo
Abstract
Feature learning with deep neural networks (DNNs) has made remarkable progress in recent years. However, its data-driven nature makes the collection of labeled training data expensive or impossible when the testing domain changes. Here, we propose a method of transferable feature learning and instance-level adaptation to improve the generalization ability of DNNs so as to mitigate the domain shift challenge for cross-domain visual recognition. When less labeled information is available, our proposed method shows attractive results in the new target domain and outperforms the typical fine-tuning method. Two DNNs are chosen as the representatives working with our proposed method, to do a comprehensive study about the generalization ability on the tasks of image-to-image transfer, image-to-video transfer, multidomain image classification, and weakly supervised detection. The experimental results show that our proposed method is superior to other existing works in the literature. In addition, a large scale of cross-domain database is merged from three different domains, providing a quantitative platform to evaluate different approaches in the field of cross-domain object detection.