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

Large-scale vision and language representation learning has shown promisingimprovements on various vision-language tasks. Most existing methods employ atransformer-based multimodal encoder to jointly model visual tokens(region-based image features) and word tokens. Because the visual tokens andword tokens are unaligned, it is challenging for the multimodal encoder tolearn image-text interactions. In this paper, we introduce a contrastive lossto ALign the image and text representations BEfore Fusing (ALBEF) them throughcross-modal attention, which enables more grounded vision and languagerepresentation learning. Unlike most existing methods, our method does notrequire bounding box annotations nor high-resolution images. In order toimprove learning from noisy web data, we propose momentum distillation, aself-training method which learns from pseudo-targets produced by a momentummodel. We provide a theoretical analysis of ALBEF from a mutual informationmaximization perspective, showing that different training tasks can beinterpreted as different ways to generate views for an image-text pair. ALBEFachieves state-of-the-art performance on multiple downstream vision-languagetasks. On image-text retrieval, ALBEF outperforms methods that are pre-trainedon orders of magnitude larger datasets. On VQA and NLVR$^2$, ALBEF achievesabsolute improvements of 2.37% and 3.84% compared to the state-of-the-art,while enjoying faster inference speed. Code and pre-trained models areavailable at https://github.com/salesforce/ALBEF/.

Align before Fuse: Vision and Language Representation Learning with Momentum Distillation