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

This paper advances the fine-grained sketch-based image retrieval (FG-SBIR)literature by putting forward a strong baseline that overshoots priorstate-of-the-arts by ~11%. This is not via complicated design though, but byaddressing two critical issues facing the community (i) the gold standardtriplet loss does not enforce holistic latent space geometry, and (ii) thereare never enough sketches to train a high accuracy model. For the former, wepropose a simple modification to the standard triplet loss, that explicitlyenforces separation amongst photos/sketch instances. For the latter, we putforward a novel knowledge distillation module can leverage photo data for modeltraining. Both modules are then plugged into a novel plug-n-playable trainingparadigm that allows for more stable training. More specifically, for (i) weemploy an intra-modal triplet loss amongst sketches to bring sketches of thesame instance closer from others, and one more amongst photos to push awaydifferent photo instances while bringing closer a structurally augmentedversion of the same photo (offering a gain of ~4-6%). To tackle (ii), we firstpre-train a teacher on the large set of unlabelled photos over theaforementioned intra-modal photo triplet loss. Then we distill the contextualsimilarity present amongst the instances in the teacher's embedding space tothat in the student's embedding space, by matching the distribution overinter-feature distances of respective samples in both embedding spaces(delivering a further gain of ~4-5%). Apart from outperforming prior artssignificantly, our model also yields satisfactory results on generalising tonew classes. Project page: https://aneeshan95.github.io/Sketch_PVT/

Exploiting Unlabelled Photos for Stronger Fine-Grained SBIR