On Fine-tuned Deep Features for Unsupervised Domain Adaptation

Wang, Q.; Meng, F.; Breckon, T.P.

Abstract

Prior feature transformation based approaches to Unsupervised Domain Adaptation (UDA) employ the deep features extracted by pre-trained deep models without fine-tuning them on the specific source or target domain data for a particular domain adaptation task. In contrast, end-to-end learning based approaches optimise the pre-trained backbones and the customised adaptation modules simultaneously to learn domaininvariant features for UDA. In this work, we explore the potential of combining fine-tuned features and feature transformation based UDA methods for improved domain adaptation performance. Specifically, we integrate the prevalent progressive pseudo-labelling techniques into the fine-tuning framework to extract fine-tuned features which are subsequently used in a state-of-the-art feature transformation based domain adaptation method SPL (Selective Pseudo-Labeling). Thorough experiments with multiple deep models including ResNet-50/101 and DeiTsmall/base are conducted to demonstrate the combination of finetuned features and SPL can achieve state-of-the-art performance on several benchmark datasets.

Citation

Wang, Q., Meng, F., & Breckon, T. (2023, June). On Fine-tuned Deep Features for Unsupervised Domain Adaptation. Presented at IJCNN 2023: International Joint Conference on Neural Networks, Queensland, Australia

• Presentation Conference Type: Conference Paper (published)
• Conference Name: IJCNN 2023: International Joint Conference on Neural Networks
• Start Date: Jun 18, 2023
• End Date: Jun 23, 2023
• Acceptance Date: Mar 30, 2023
• Online Publication Date: Aug 2, 2023
• Publication Date: 2023-08
• Deposit Date: Jun 2, 2023
• Publicly Available Date: Sep 13, 2023
• Publisher: Institute of Electrical and Electronics Engineers
• Series Title: IJCNN 2023 Conference Proceedings
• Series ISSN: 2161-4407
• DOI: https://doi.org/10.1109/IJCNN54540.2023.10191262
• Public URL: https://durham-repository.worktribe.com/output/1134732
• Publisher URL: https://ieeexplore.ieee.org/xpl/conhome/1000500/all-proceedings