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

Recent works in self-supervised learning have shown impressive results onsingle-object images, but they struggle to perform well on complex multi-objectimages as evidenced by their poor visual grounding. To demonstrate thisconcretely, we propose visual difference attention (VDA) to compute visualattention maps in an unsupervised fashion by comparing an image with itssalient-regions-masked-out version. We use VDA to derive attention maps forstate-of-the art SSL methods and show they do not highlight all salient regionsin an image accurately, suggesting their inability to learn strongrepresentations for downstream tasks like segmentation. Motivated by theselimitations, we cast VDA as a differentiable operation and propose a newlearning objective, Differentiable Difference Attention (DiDA) loss, whichleads to substantial improvements in an SSL model's visually grounding to animage's salient regions.

Learning with Difference Attention for Visually Grounded Self-supervised Representations