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

More information leads to better decisions and predictions, right? Confirmingthis hypothesis, several studies concluded that the simultaneous use of opticaland thermal images leads to better predictions in crowd counting. However, theway multimodal models extract enriched features from both modalities is not yetfully understood. Since the use of multimodal data usually increases thecomplexity, inference time, and memory requirements of the models, it isrelevant to examine the differences and advantages of multimodal compared tomonomodal models. In this work, all available multimodal datasets for crowdcounting are used to investigate the differences between monomodal andmultimodal models. To do so, we designed a monomodal architecture thatconsiders the current state of research on monomodal crowd counting. Inaddition, several multimodal architectures have been developed using differentmultimodal learning strategies. The key components of the monomodalarchitecture are also used in the multimodal architectures to be able to answerwhether multimodal models perform better in crowd counting in general.Surprisingly, no general answer to this question can be derived from theexisting datasets. We found that the existing datasets hold a bias towardthermal images. This was determined by analyzing the relationship between thebrightness of optical images and crowd count as well as examining theannotations made for each dataset. Since answering this question is importantfor future real-world applications of crowd counting, this paper establishescriteria for a potential dataset suitable for answering whether multimodalmodels perform better in crowd counting in general.

Why Existing Multimodal Crowd Counting Datasets Can Lead to Unfulfilled Expectations in Real-World Applications