图像数据处理模型训练系统设计与实现
Design and Implementation of Image Data Processing Model Training System
杨政 1徐亮 1廖一星 1王亮 1官永杨1
作者信息
- 1. 贵州江南航天信息网络通信有限公司,贵州 遵义 563000
- 折叠
摘要
结合复杂图像数据处理及模型训练应用场景需求,开展图像数据处理系统模块化设计与搭建工作,将图像处理功能单元及模型库进行封装处理,同时搭建训练及接口并通过构建API进行功能和处理模型的调度,并实现功能单元和模型快速组合与应用,以此解决图像数据处理耗时长、处理工作量大以及图像数据的模型训练不及时等问题,为机器视觉技术在产品外观缺陷识别、缺陷目标特征定位及检测等应用方面提供理论支撑和技术参考.
Abstract
This paper combines the requirements of complex image data processing and model training application sce-narios to carry out modular design and construction of image data processing systems,encapsulating image processing functional units and model libraries.Simultaneously building training and interfaces,and scheduling functions and processing models through API construction,and achieving rapid combination and application of functional units and models,in order to solve problems such as long processing time,large processing workload,and delayed model training of image data.This provides theoretical support and technical reference for machine vision technology in product appearance defect recognition,defect target feature localization,and detection applications.
关键词
图像数据/模型训练/模块化设计Key words
image data/model training/modular design引用本文复制引用
基金项目
贵州省科技支撑计划(黔科合支撑[2023]一般323)
遵义市科技创新人才团队培养计划(遵市科人才[2022]6号)
出版年
2024
NETL
NSTL
万方数据