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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Vision-Language Pre-Training (VLP) has shown promising capabilities to alignimage and text pairs, facilitating a broad variety of cross-modal learningtasks. However, we observe that VLP models often lack the visualgrounding/localization capability which is critical for many downstream taskssuch as visual reasoning. In this work, we propose a novel Position-guided TextPrompt (PTP) paradigm to enhance the visual grounding ability of cross-modalmodels trained with VLP. Specifically, in the VLP phase, PTP divides the imageinto $N\times N$ blocks, and identifies the objects in each block through thewidely used object detector in VLP. It then reformulates the visual groundingtask into a fill-in-the-blank problem given a PTP by encouraging the model topredict the objects in the given blocks or regress the blocks of a givenobject, e.g. filling `P" or ``O" in aPTP ``The block P has a O". This mechanismimproves the visual grounding capability of VLP models and thus helps thembetter handle various downstream tasks. By introducing PTP into severalstate-of-the-art VLP frameworks, we observe consistently significantimprovements across representative cross-modal learning model architectures andseveral benchmarks, e.g. zero-shot Flickr30K Retrieval (+4.8 in averagerecall@1) for ViLT \cite{vilt} baseline, and COCO Captioning (+5.3 in CIDEr)for SOTA BLIP \cite{blip} baseline. Moreover, PTP achieves comparable resultswith object-detector based methods, and much faster inference speed since PTPdiscards its object detector for inference while the later cannot. Our code andpre-trained weight will be released at \url{https://github.com/sail-sg/ptp}.

Position-guided Text Prompt for Vision-Language Pre-training