基于改进HRNet的牛体关键点检测算法
Cattle body keypoint detection algorithm based on improved HRNet
赵雪莲 1张继凯 1何一豪 1曾翔皓 1庄琦2
作者信息
- 1. 内蒙古科技大学信息工程学院,内蒙古包头 014010
- 2. 内蒙古科技大学工程训练中心,内蒙古包头 014010
- 折叠
摘要
针对现有关键点检测算法在复杂背景下检测精度低、高运算量等问题,提出一种轻量级关键点检测模型SE-HRNet.首先设计2 种轻量型模块:SECAneck模块和SECAblock模块,在保持网络性能的同时减低计算参数,加快训练速度.其次,整合空间注意力机制于多分辨率融合阶段,使得模型对于不易检测到的关键点的定位和识别更为敏感.在自制牛体关键点数据集上进行实验评估,结果表明:改进后的HRNet网络比原网络参数量和运算浮点数分别减少了18.8 M和5.2 G,平均精度达到了 93.2%,平均召回率达到了 91.5%,每秒帧数(FPS)达到了36.3.
Abstract
In response to the low accuracy and high computational demands of existing keypoint detection algorithms in complex envi-ronments,the SE-HRNet,a lightweight model was developed.This was accomplished by first creating two lightweight modules,the SECAneck and the SECAblock,which allowed for a reduction in computational parameters and faster training without sacrificing per-formance.In addition,a spatial attention mechanism was incorporated with the multi-resolution fusion stage,improving the model's sensitivity to difficult-to-detect keypoints.Experiments on custom-built cattle body keypoint dataset show that the refined HRNet net-work decreases parameter count by 18.8 million and computational operations by 5.2 billion.The model now achieves an average preci-sion of 93.2%and the average recall rate of 91.5%,while running at 36.3 frames per second.
关键词
关键点检测/高分辨率网络/注意力机制/多分辨率融合阶段Key words
keypoint detection/high-resolution networks/attention mechanism/multi-resolution fusion stage引用本文复制引用
基金项目
内蒙古自治区科技重大专项(2019ZD025)
内蒙古自治区自然科学基金(2021MS06007)
内蒙古自治区科技计划项目(2019GG138)
出版年
2024
NETL
NSTL
万方数据