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/

Zendy Plus

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

Zendy Tools

Zendy Open

Recently, significant advancements in artificial intelligence have beenattributed to the integration of self-supervised learning (SSL) scheme. WhileSSL has shown impressive achievements in natural language processing (NLP), itsprogress in computer vision has comparatively lagged behind. However, theincorporation of contrastive learning into existing visual SSL models has ledto considerable progress, often surpassing supervised counterparts.Nonetheless, these improvements have been mostly limited to classificationtasks. Moreover, few studies have evaluated visual SSL models in real-worldscenarios, as most have focused on datasets with class-wise portrait images,notably ImageNet. Here, we focus on dense prediction tasks using securityinspection x-ray images to evaluate our proposed model, Segment Localization(SegLoc). Based upon the Instance Localization (InsLoc) model, SegLoc addressesone of the key challenges of contrastive learning, i.e., false negative pairsof query embeddings. Our pre-training dataset is synthesized by cutting,transforming, and pasting labeled segments from an existing labeled dataset(PIDray) as foregrounds onto instances from an unlabeled dataset (SIXray) asbackgrounds. Furthermore, we fully leverage the labeled data by incorporatingthe concept, one queue per class, into the MoCo-v2 memory bank, therebyavoiding false negative pairs. In our experiments, SegLoc outperformed randominitialization by 3% to 6% while underperformed supervised initialization, interms of AR and AP metrics across different IoU values over 20 to 30pre-training epochs.

Visual Self-supervised Learning Scheme for Dense Prediction Tasks on X-ray Images