包装箱是商品流通过程中的重要保障之一,针对跨国运输的外贸工艺品,确保外包装纸箱外观整洁无破损、变形和污渍是极其必要的.文章提出了一种全新的包装箱缺陷检测解决方案,将目标检测技术应用于国内外贸物流仓库;通过在物流仓库进行实际考察拍摄,自制数据集,分别训练YOLOx、YOLOv4、YOLOv5 模型.通过检测结果对比得出,YOLOv5 为检测效果最准确、效率最高的模型,平均精度均值(Mean Average Precision,mAP)为 90.3%,平均精度(Average Precision,AP)值最大,每秒检测帧数(Frames Per Second,FPS)值可满足实际应用需求.
Craft box defect detection based on YOLOv5
Packaging boxes are one of the important guarantees in the circulation process of goods,for the foreign trade handicrafts transported across borders,it is extremely necessary to ensure that the appearance of the outer packaging box is clean and free of damage,deformation and stains.This article proposes a brand-new packaging box defect detection solution,applying target detection technology to domestic and foreign trade logistics warehouses for the first time.By photographing through actual inspections in logistics warehouses,a dataset is made,and YOLOx,YOLOv4,YOLOv5 models are respectively trained.By comparing the test results,it can be concluded that YOLOv5 is the most accurate and efficient model for detection.Its mean average precision is 90.3%,average precision value is max,the frames per second value can meet actual application needs.
NETL
NSTL
万方数据
