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

Multimodal alignment between language and vision is the fundamental topic incurrent vision-language model research. Contrastive Captioners (CoCa), as arepresentative method, integrates Contrastive Language-Image Pretraining (CLIP)and Image Caption (IC) into a unified framework, resulting in impressiveresults. CLIP imposes a bidirectional constraints on global representation ofentire images and sentences. Although IC conducts an unidirectionalimage-to-text generation on local representation, it lacks any constraint onlocal text-to-image reconstruction, which limits the ability to understandimages at a fine-grained level when aligned with texts. To achieve multimodalalignment from both global and local perspectives, this paper proposesSymmetrizing Contrastive Captioners (SyCoCa), which introduces bidirectionalinteractions on images and texts across the global and local representationlevels. Specifically, we expand a Text-Guided Masked Image Modeling (TG-MIM)head based on ITC and IC heads. The improved SyCoCa can further leveragetextual cues to reconstruct contextual images and visual cues to predicttextual contents. When implementing bidirectional local interactions, the localcontents of images tend to be cluttered or unrelated to their textualdescriptions. Thus, we employ an attentive masking strategy to select effectiveimage patches for interaction. Extensive experiments on five vision-languagetasks, including image-text retrieval, image-captioning, visual questionanswering, and zero-shot/finetuned image classification, validate theeffectiveness of our proposed method.

SyCoCa: Symmetrizing Contrastive Captioners with Attentive Masking for Multimodal Alignment