The network architecture of YOLOv8 comprises three parts.
figure
posted on 2025-05-07, 17:36 authored by Xunqian Xu, Tao Wu, Zhongbao Du, Hui Rong, Siwen Wang, Shue Li, Dakai ChenThe Backbone is responsible for feature extraction. The Neck, situated between the Backbone and the Head, is responsible for feature fusion. The Head is responsible for outputting the detection results.
History
Usage metrics
Categories
Keywords
novel attention mechanismhuman body behavioraverage frame rateapproximately 3 frames3 &# 8211new loss functionmodel &# 8217pose estimation datasetoriginal bottleneck modulefeature fusion moduleexperimental results demonstratetraditional detection algorithmsposture human motion analysisyolov8 loss functionoriginal yolov8 modeldciou loss functionpose detectionloss functionstraditional yolov8model trainingresults indicateyolov8 frameworksimdlka modulevaluable referencethereby improvingsource cocosignificant interestselectively focusmpii datasetsmethod providesmap valueinput dataindependently constructedforward propagationdetecting posescrucial taskc2f modulealso suffer
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC