Fall Detection in Multiple Scenarios Based on YOLOv8
PIAN Lulu
PEI Huandou
ZHANG Yuxuan
Abstract：In order to solve the problems of poor accuracy,slow detection speed and single detection background of fall detec-tion algorithm,an improved YOLOv8 algorithm for multiple background environments is proposed.The neck structure of the bench-mark model is replaced by Bic FPN-PAN structure,and the detection accuracy is improved by the reuse of the same layer feature in-formation and the bottom detail information.The addition operation is used for feature fusion to ensure the lightweight of the model.Replacing the C2f module in the backbone of the benchmark model with GS2C module can reduce the false detection rate and the number of parameters in the model.After experiments,the accuracy of the improved model is increased by 5.27%,mAP is in-creased by 2.83%,model computation is reduced by 1%.Compared with the current mainstream YOLO series algorithm,the im-proved algorithm has a better effect.
Keywords：deep learningfall detectionattention mechanismfeature fusion
Publication Date：2025-11-20
Online Publishing Date：2026-01-28(First online date of this platform, not the publication date of the document)
Pages：5( 56-60 )