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

As an emerging topic in face recognition, designing margin-based lossfunctions can increase the feature margin between different classes forenhanced discriminability. More recently, the idea of mining-based strategiesis adopted to emphasize the misclassified samples, achieving promising results.However, during the entire training process, the prior methods either do notexplicitly emphasize the sample based on its importance that renders the hardsamples not fully exploited; or explicitly emphasize the effects ofsemi-hard/hard samples even at the early training stage that may lead toconvergence issue. In this work, we propose a novel Adaptive CurriculumLearning loss (CurricularFace) that embeds the idea of curriculum learning intothe loss function to achieve a novel training strategy for deep facerecognition, which mainly addresses easy samples in the early training stageand hard ones in the later stage. Specifically, our CurricularFace adaptivelyadjusts the relative importance of easy and hard samples during differenttraining stages. In each stage, different samples are assigned with differentimportance according to their corresponding difficultness. Extensiveexperimental results on popular benchmarks demonstrate the superiority of ourCurricularFace over the state-of-the-art competitors.

CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition