改进YOLOv3的行人车辆目标检测算法

An Improved Algorithm of Pedestrian and Vehicle Detection Based on YOLOv3

针对YOLOv3(you only look once version 3)对中小目标检测效果不理想的问题,提出改进算法DX-YOLO(densely ResneXt with YOLOv3)。首先对YOLOv3的特征提取网络Darknet-53进行改进,使用ResneXt残差模块替换原有残差模块,优化了卷积网络结构;受DenseNet的启发,在Darknet-53中引入密集连接,实现了特征重用,提高了提取特征的效率;根据数据集的特点,利用K-means算法对数据集进行维度聚类,获得合适的预选框。在行人车辆数据集Udacity上进行实验,结果表明:DX-YOLO算法与YOLOv3相比,平均准确率(mean average precision,mAP)提升了3.42%;特别地,在中等目标和小目标上的平均精度(average precision,AP)分别提升了2.74%和5.98%。

Considering that YOLOv3(you only look once version 3)was not ideal for small or medium targets detection,an improved algorithm DX-YOLO(densely ResneXt with YOLOv3)was proposed.Firstly,the feature extraction network of YOLOv3 called Darknet-53 was improved.Then,the original residual module was replaced by ResneXt residual module,which optimized the structure of convolution network.Inspired by DenseNet,dense connection was introduced into Darknet-53 to realize feature reuse,and improved the efficiency of feature extraction.According to the characteristics of data set,K-means algorithm was used to cluster the dimensions of data set to get the appropriate anchor box.Experiments on Udacity data set show that compared with YOLOv3,DX-YOLO algorithm improves the mean average precision mean average precision(mAP)by 3.42%.Especially,the average precision(AP)on medium and small targets increases by 2.74%and 5.98% respectively.