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

Crowd counting is an application-oriented task and its inference efficiencyis crucial for real-world applications. However, most previous works relied onheavy backbone networks and required prohibitive run-time consumption, whichwould seriously restrict their deployment scopes and cause poor scalability. Toliberate these crowd counting models, we propose a novel Structured KnowledgeTransfer (SKT) framework, which fully exploits the structured knowledge of awell-trained teacher network to generate a lightweight but still highlyeffective student network. Specifically, it is integrated with twocomplementary transfer modules, including an Intra-Layer Pattern Transfer whichsequentially distills the knowledge embedded in layer-wise features of theteacher network to guide feature learning of the student network and anInter-Layer Relation Transfer which densely distills the cross-layercorrelation knowledge of the teacher to regularize the student's featureevolutio Consequently, our student network can derive the layer-wise andcross-layer knowledge from the teacher network to learn compact yet effectivefeatures. Extensive evaluations on three benchmarks well demonstrate theeffectiveness of our SKT for extensive crowd counting models. In particular,only using around $6\%$ of the parameters and computation cost of originalmodels, our distilled VGG-based models obtain at least 6.5$\times$ speed-up onan Nvidia 1080 GPU and even achieve state-of-the-art performance. Our code andmodels are available at {\url{https://github.com/HCPLab-SYSU/SKT}}.

Efficient Crowd Counting via Structured Knowledge Transfer