基于改进YOLOv8s的密集多人脸检测

A Dense Multi-face Detection Based on Improved YOLOv8s

针对密集场景中多人脸检测存在的漏检率高、检测精度低的问题,提出一种改进的YOLOv8s人脸检测算法。在YOLOv8的骨干网络添加SimAM注意力机制,提高检测模型对于图像内小目标的特征提取能力。将原激活函数SiLU替换为FReLU函数,扩大特征点提取范围,提高小目标检测准确度。引入新的损失函数Wise-IoUv1,解决部分小目标在截取过程中可能出现的低质量问题,进一步提升检测精度。实验结果表明,改进后的算法在自建密集场景人脸数据集上,准确度提升到99.26%,在回归率持平的基础上计算参数无明显上涨,实测漏检率降低26%,有效提升了人脸检测能力。

Aiming at the problems of high missed detection rate and low detection accuracy in dense multiface detection,an improved face detection algorithm based on YOLOv8s is proposed.The SimAM attention mechanism is incorporated into the backbone network of YOLOv8 to enhance the feature extraction ability of the detection model for small targets in the images.The original SiLU activation function is replaced with the FReLU function to expand the range of feature point extraction and improve the accuracy of small object detection.A new loss function Wise-IoUv1 is introduced to solve the low-quality problem that may occur during the interception process of some small targets,and further improve the detection accuracy.The experimental results show that the improved algorithm can achieve an accuracy improvement up to 99.26%on the self-built face data set in dense background without significant increase in computational parameters compared with flat regression rates.It can reduce 26%missed detection rate,enhancing face detection capability effectively.