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

Contrastive self-supervised learning (CSL) has attracted increasing attentionfor model pre-training via unlabeled data. The resulted CSL models provideinstance-discriminative visual features that are uniformly scattered in thefeature space. During deployment, the common practice is to directly fine-tuneCSL models with cross-entropy, which however may not be the best strategy inpractice. Although cross-entropy tends to separate inter-class features, theresulting models still have limited capability for reducing intra-class featurescattering that exists in CSL models. In this paper, we investigate whetherapplying contrastive learning to fine-tuning would bring further benefits, andanalytically find that optimizing the contrastive loss benefits bothdiscriminative representation learning and model optimization duringfine-tuning. Inspired by these findings, we propose Contrast-regularized tuning(Core-tuning), a new approach for fine-tuning CSL models. Instead of simplyadding the contrastive loss to the objective of fine-tuning, Core-tuningfurther applies a novel hard pair mining strategy for more effectivecontrastive fine-tuning, as well as smoothing the decision boundary to betterexploit the learned discriminative feature space. Extensive experiments onimage classification and semantic segmentation verify the effectiveness ofCore-tuning.

Unleashing the Power of Contrastive Self-Supervised Visual Models via Contrast-Regularized Fine-Tuning