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

Panoptic Narrative Grounding (PNG) is an emerging task whose goal is tosegment visual objects of things and stuff categories described by densenarrative captions of a still image. The previous two-stage approach firstextracts segmentation region proposals by an off-the-shelf panopticsegmentation model, then conducts coarse region-phrase matching to ground thecandidate regions for each noun phrase. However, the two-stage pipeline usuallysuffers from the performance limitation of low-quality proposals in the firststage and the loss of spatial details caused by region feature pooling, as wellas complicated strategies designed for things and stuff categories separately.To alleviate these drawbacks, we propose a one-stage end-to-end Pixel-PhraseMatching Network (PPMN), which directly matches each phrase to itscorresponding pixels instead of region proposals and outputs panopticsegmentation by simple combination. Thus, our model can exploit sufficient andfiner cross-modal semantic correspondence from the supervision of denselyannotated pixel-phrase pairs rather than sparse region-phrase pairs. Inaddition, we also propose a Language-Compatible Pixel Aggregation (LCPA) moduleto further enhance the discriminative ability of phrase features throughmulti-round refinement, which selects the most compatible pixels for eachphrase to adaptively aggregate the corresponding visual context. Extensiveexperiments show that our method achieves new state-of-the-art performance onthe PNG benchmark with 4.0 absolute Average Recall gains.

PPMN: Pixel-Phrase Matching Network for One-Stage Panoptic Narrative Grounding