Spectral Unsupervised Domain Adaptation for Visual Recognition
This addresses the problem of adapting models across domains without target labels for visual recognition tasks, offering a novel method that is incremental in improving existing UDA techniques.
The paper tackles unsupervised domain adaptation for visual recognition by proposing Spectral UDA (SUDA), which uses a spectrum transformer and multi-view spectral learning to reduce domain discrepancies and learn unsupervised representations, achieving superior accuracy across tasks like object detection, semantic segmentation, and image classification, with state-of-the-art performance on object detection using transformer-based networks.
Though unsupervised domain adaptation (UDA) has achieved very impressive progress recently, it remains a great challenge due to missing target annotations and the rich discrepancy between source and target distributions. We propose Spectral UDA (SUDA), an effective and efficient UDA technique that works in the spectral space and can generalize across different visual recognition tasks. SUDA addresses the UDA challenges from two perspectives. First, it introduces a spectrum transformer (ST) that mitigates inter-domain discrepancies by enhancing domain-invariant spectra while suppressing domain-variant spectra of source and target samples simultaneously. Second, it introduces multi-view spectral learning that learns useful unsupervised representations by maximizing mutual information among multiple ST-generated spectral views of each target sample. Extensive experiments show that SUDA achieves superior accuracy consistently across different visual tasks in object detection, semantic segmentation and image classification. Additionally, SUDA also works with the transformer-based network and achieves state-of-the-art performance on object detection.