福建师范大学学报（自然科学版）2025，Vol.41Issue(3)：19-26,8.DOI:10.12046/j.issn.1000-5277.2024010029
一种基于特征融合的轻量级姿态估计算法
A Lightweight Pose Estimation Algorithm Based on Feature Fusion
摘要
Abstract
In human pose estimation tasks,most existing studies primarily focus on the accu-racy of models while overlooking efficiency-related factors such as model size,parameter count,and inference time.However,these metrics are crucial for practical applications.To address this is-sue,this paper proposes a lightweight human pose estimation algorithm based on the YOLOv8-pose algorithm.The algorithm incorporates a cross-scale feature fusion module(CCFM)to enhance the model's adaptability to scale variations and its detection capability for small-scale objects.By effec-tively combining detailed features and contextual information,the model's overall performance is im-proved,and its parameter count is reduced.Additionally,SENetV2 is used to replace the convolu-tional structure in the C2f module of YOLOv8-pose,strengthening the model's global consideration and improving its prediction accuracy.The adoption of the MPDIoU loss function further improves the model's calculation of bounding box errors during training,thereby boosting inference accuracy.The proposed approach achieves a 1.3%improvement in mAP50∶95 compared to the original model.关键词
姿态估计/YOLOv8/轻量化/特征融合Key words
pose estimation/YOLOv8/lightweight/feature fusion分类
计算机与自动化引用本文复制引用
张伟杰,叶锋,李昊珑..一种基于特征融合的轻量级姿态估计算法[J].福建师范大学学报（自然科学版）,2025,41(3):19-26,8.基金项目
福建省科技创新战略研究联合计划项目(2023R0156) （2023R0156）
