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

Most previous approaches for analyzing food images have relied on extensivelyannotated datasets, resulting in significant human labeling expenses due to thevaried and intricate nature of such images. Inspired by the effectiveness ofcontrastive self-supervised methods in utilizing unlabelled data, we exploreleveraging these techniques on unlabelled food images. In contrastiveself-supervised methods, two views are randomly generated from an image by dataaugmentations. However, regarding food images, the two views tend to containsimilar informative contents, causing large mutual information, which impedesthe efficacy of contrastive self-supervised learning. To address this problem,we propose Feature Suppressed Contrast (FeaSC) to reduce mutual informationbetween views. As the similar contents of the two views are salient or highlyresponsive in the feature map, the proposed FeaSC uses a response-aware schemeto localize salient features in an unsupervised manner. By suppressing somesalient features in one view while leaving another contrast view unchanged, themutual information between the two views is reduced, thereby enhancing theeffectiveness of contrast learning for self-supervised food pre-training. As aplug-and-play module, the proposed method consistently improves BYOL andSimSiam by 1.70\% $\sim$ 6.69\% classification accuracy on four publiclyavailable food recognition datasets. Superior results have also been achievedon downstream segmentation tasks, demonstrating the effectiveness of theproposed method.

Feature-Suppressed Contrast for Self-Supervised Food Pre-training