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

Tremendous progress has been made in visual representation learning, notablywith the recent success of self-supervised contrastive learning methods.Supervised contrastive learning has also been shown to outperform itscross-entropy counterparts by leveraging labels for choosing where to contrast.However, there has been little work to explore the transfer capability ofcontrastive learning to a different domain. In this paper, we conduct acomprehensive study on the transferability of learned representations ofdifferent contrastive approaches for linear evaluation, full-network transfer,and few-shot recognition on 12 downstream datasets from different domains, andobject detection tasks on MSCOCO and VOC0712. The results show that thecontrastive approaches learn representations that are easily transferable to adifferent downstream task. We further observe that the joint objective ofself-supervised contrastive loss with cross-entropy/supervised-contrastive lossleads to better transferability of these models over their supervisedcounterparts. Our analysis reveals that the representations learned from thecontrastive approaches contain more low/mid-level semantics than cross-entropymodels, which enables them to quickly adapt to a new task. Our codes and modelswill be publicly available to facilitate future research on transferability ofvisual representations.

A Broad Study on the Transferability of Visual Representations with Contrastive Learning