English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Detection transformers have recently shown promising object detection resultsand attracted increasing attention. However, how to develop effective domainadaptation techniques to improve its cross-domain performance remainsunexplored and unclear. In this paper, we delve into this topic and empiricallyfind that direct feature distribution alignment on the CNN backbone only bringslimited improvements, as it does not guarantee domain-invariant sequencefeatures in the transformer for prediction. To address this issue, we propose anovel Sequence Feature Alignment (SFA) method that is specially designed forthe adaptation of detection transformers. Technically, SFA consists of a domainquery-based feature alignment (DQFA) module and a token-wise feature alignment(TDA) module. In DQFA, a novel domain query is used to aggregate and alignglobal context from the token sequence of both domains. DQFA reduces the domaindiscrepancy in global feature representations and object relations whendeploying in the transformer encoder and decoder, respectively. Meanwhile, TDAaligns token features in the sequence from both domains, which reduces thedomain gaps in local and instance-level feature representations in thetransformer encoder and decoder, respectively. Besides, a novel bipartitematching consistency loss is proposed to enhance the feature discriminabilityfor robust object detection. Experiments on three challenging benchmarks showthat SFA outperforms state-of-the-art domain adaptive object detection methods.Code has been made available at: https://github.com/encounter1997/SFA.

Exploring Sequence Feature Alignment for Domain Adaptive Detection Transformers