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

Webly supervised learning becomes attractive recently for its efficiency indata expansion without expensive human labeling. However, adopting searchqueries or hashtags as web labels of images for training brings massive noisethat degrades the performance of DNNs. Especially, due to the semanticconfusion of query words, the images retrieved by one query may containtremendous images belonging to other concepts. For example, searching `tigercat' on Flickr will return a dominating number of tiger images rather than thecat images. These realistic noisy samples usually have clear visual semanticclusters in the visual space that mislead DNNs from learning accurate semanticlabels. To correct real-world noisy labels, expensive human annotations seemindispensable. Fortunately, we find that metadata can provide extra knowledgeto discover clean web labels in a labor-free fashion, making it feasible toautomatically provide correct semantic guidance among the massive label-noisyweb data. In this paper, we propose an automatic label corrector VSGraph-LCbased on the visual-semantic graph. VSGraph-LC starts from anchor selectionreferring to the semantic similarity between metadata and correct labelconcepts, and then propagates correct labels from anchors on a visual graphusing graph neural network (GNN). Experiments on realistic webly supervisedlearning datasets Webvision-1000 and NUS-81-Web show the effectiveness androbustness of VSGraph-LC. Moreover, VSGraph-LC reveals its advantage on theopen-set validation set.

Webly Supervised Image Classification with Metadata: Automatic Noisy Label Correction via Visual-Semantic Graph