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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Existing long-tailed recognition methods, aiming to train class-balancedmodels from long-tailed data, generally assume the models would be evaluated onthe uniform test class distribution. However, practical test classdistributions often violate this assumption (e.g., being either long-tailed oreven inversely long-tailed), which may lead existing methods to fail in realapplications. In this paper, we study a more practical yet challenging task,called test-agnostic long-tailed recognition, where the training classdistribution is long-tailed while the test class distribution is agnostic andnot necessarily uniform. In addition to the issue of class imbalance, this taskposes another challenge: the class distribution shift between the training andtest data is unknown. To tackle this task, we propose a novel approach, calledSelf-supervised Aggregation of Diverse Experts, which consists of twostrategies: (i) a new skill-diverse expert learning strategy that trainsmultiple experts from a single and stationary long-tailed dataset to separatelyhandle different class distributions; (ii) a novel test-time expert aggregationstrategy that leverages self-supervision to aggregate the learned multipleexperts for handling unknown test class distributions. We theoretically showthat our self-supervised strategy has a provable ability to simulatetest-agnostic class distributions. Promising empirical results demonstrate theeffectiveness of our method on both vanilla and test-agnostic long-tailedrecognition. Code is available at\url{https://github.com/Vanint/SADE-AgnosticLT}.

Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition