Leather defect detection method, system and device based on YOLOv5

Abstract

The invention provides a leather defect detection method, system and device based on YOLOv5, and solves the problems of single detection mode, high dependency on a deep learning network, difficulty in adapting to leather defect variability and uncertain production environment in an actual leather production process in a current leather defect detection method by utilizing a traditional deep learning network. According to the invention, firstly, the leather defect detection network model based on YOLOv5 is constructed, the detection speed is high, and the accuracy is high; then the leather defect detection network model is deployed by adopting a dual-mode deployment design of a GPU server deployment mode based on GRPC and an offline edge calculation deployment mode, so as to adapt to defect variability in the leather production process, and the requirements of different leather production environments are met.

Description

technical field [0001] The present invention relates to the technical field of leather defect detection, and more specifically, to a method, system and device for detecting leather defects based on YOLOv5. Background technique [0002] Leather is widely used in clothing, bags, decorative accessories and other daily items. With the increasing demand for leather and the continuous improvement of consumers' requirements for the quality of leather products, leather factories are also stricter in controlling the quality of leather. However, leather Products are susceptible to various defects such as pinholes, air bubbles, color lines, dirt, etc. due to the original source environment or manufacturing process. [0003] At present, most of the means of leather defect detection remain at the level of offline manual detection. On the one hand, the heavy detection work and the harsh environment such as high noise, dim light, and turbid smell on the production line are extremely harmfu...

AI Technical Summary

Problems solved by technology

[0005]In order to solve the current method of using the traditional deep learning network for leather defect detection, the detection mode is single, and it is highly dependent on the deep learning network, which is difficult to adapt to the actual leather production process. To solve the problems of defect variability and uncertain production environment, the present invention proposes a leather defect detection method, system and device based on YOLOv5 to realize incremental learning of leather defects, improve detection efficiency, and adapt to defect variability in the leather production process

Images