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

Currently, for crowd counting, the fully supervised methods via density mapestimation are the mainstream research directions. However, such methods needlocation-level annotation of persons in an image, which is time-consuming andlaborious. Therefore, the weakly supervised method just relying upon thecount-level annotation is urgently needed. Since CNN is not suitable formodeling the global context and the interactions between image patches, crowdcounting with weakly supervised learning via CNN generally can not show goodperformance. The weakly supervised model via Transformer was sequentiallyproposed to model the global context and learn contrast features. However, thetransformer directly partitions the crowd images into a series of tokens, whichmay not be a good choice due to each pedestrian being an independentindividual, and the parameter number of the network is very large. Hence, wepropose a Joint CNN and Transformer Network (JCTNet) via weakly supervisedlearning for crowd counting in this paper. JCTNet consists of three parts: CNNfeature extraction module (CFM), Transformer feature extraction module (TFM),and counting regression module (CRM). In particular, the CFM extracts crowdsemantic information features, then sends their patch partitions to TRM formodeling global context, and CRM is used to predict the number of people.Extensive experiments and visualizations demonstrate that JCTNet caneffectively focus on the crowd regions and obtain superior weakly supervisedcounting performance on five mainstream datasets. The number of parameters ofthe model can be reduced by about 67%~73% compared with the pure Transformerworks. We also tried to explain the phenomenon that a model constrained only bycount-level annotations can still focus on the crowd regions. We believe ourwork can promote further research in this field.

Joint CNN and Transformer Network via weakly supervised Learning for efficient crowd counting