View in EDS

A Crowd Counting and Localization Network Based on Adaptive Feature Fusion and Multi-Scale Global Attention Up Sampling

Saved in:
Bibliographic Details
Title: A Crowd Counting and Localization Network Based on Adaptive Feature Fusion and Multi-Scale Global Attention Up Sampling
Authors: Min Wang, Li Huang, Jingke Yan, Jin Huang, Tao Yang
Source: IEEE Access, Vol 12, Pp 12919-12939 (2024)
Publisher Information: Institute of Electrical and Electronics Engineers (IEEE), 2024.
Publication Year: 2024
Subject Terms: global attention, adaptive feature fusion, Label map, multi-scale, upsampling, 0202 electrical engineering, electronic engineering, information engineering, 0401 agriculture, forestry, and fisheries, encoder-decoder structure, Electrical engineering. Electronics. Nuclear engineering, 04 agricultural and veterinary sciences, 02 engineering and technology, TK1-9971
Description: Crowd counting is an important research topic in the fields of computer vision and image processing, with monitoring and management of crowded scenes becoming an increasingly prominent issue. Existing methods still suffer from the problem of severe overlap in density maps within dense areas, leading to inadequate counting and localization accuracy. This paper presents innovative research on crowd counting and localization. Firstly, addressing the limitations of density maps in localization performance in existing algorithms, we optimize the generation method of FIDT maps, decoupling the counting and localization tasks. By avoiding the problem of overlap in dense areas, the optimized label maps achieve a good balance between counting accuracy and localization, with MAE and MSE reaching 64.1 and 103.9 in SHHA, and 10.9 and 17.4 in SHHB, respectively.Secondly, to address the scale insensitivity of the encoder and the potential loss of critical features during the encoding process, we propose the Adaptive Feature Fusion Module and the Multi-Scale Global Attention Upsampling Module, constructing the CALNET network. By reducing redundant features inside and outside the separable branch, the model achieves global fusion of shallow features during the decoding process. The F1-m scores obtained on the SHHA and SHHB datasets reach 72.9% and 79.4% respectively, significantly improving the model’s performance.Finally, this paper extends the application of crowd counting and localization algorithms to different domains such as citrus orchards, vehicles, and campus crowds. Through experiments, the robustness and transferability of the network are validated, expanding the application areas of crowd counting and localization algorithms and providing a broader space for future research.
Document Type: Article
ISSN: 2169-3536
DOI: 10.1109/access.2024.3356604
Access URL: https://doaj.org/article/73ce4e74c5b04daca73bde32c0c8842d
Rights: CC BY NC ND
Accession Number: edsair.doi.dedup.....5a9536e0bb44a85f95b3daed55372bfc
Database: OpenAIRE
Description
Abstract:Crowd counting is an important research topic in the fields of computer vision and image processing, with monitoring and management of crowded scenes becoming an increasingly prominent issue. Existing methods still suffer from the problem of severe overlap in density maps within dense areas, leading to inadequate counting and localization accuracy. This paper presents innovative research on crowd counting and localization. Firstly, addressing the limitations of density maps in localization performance in existing algorithms, we optimize the generation method of FIDT maps, decoupling the counting and localization tasks. By avoiding the problem of overlap in dense areas, the optimized label maps achieve a good balance between counting accuracy and localization, with MAE and MSE reaching 64.1 and 103.9 in SHHA, and 10.9 and 17.4 in SHHB, respectively.Secondly, to address the scale insensitivity of the encoder and the potential loss of critical features during the encoding process, we propose the Adaptive Feature Fusion Module and the Multi-Scale Global Attention Upsampling Module, constructing the CALNET network. By reducing redundant features inside and outside the separable branch, the model achieves global fusion of shallow features during the decoding process. The F1-m scores obtained on the SHHA and SHHB datasets reach 72.9% and 79.4% respectively, significantly improving the model’s performance.Finally, this paper extends the application of crowd counting and localization algorithms to different domains such as citrus orchards, vehicles, and campus crowds. Through experiments, the robustness and transferability of the network are validated, expanding the application areas of crowd counting and localization algorithms and providing a broader space for future research.
ISSN:21693536
DOI:10.1109/access.2024.3356604