The integration of automation equipment with intelligence has undoubtedly become the trend in the automation industry. In this regard, automated optical inspection (AOI) application, with advantages such as not being limited by working hours, rapid inspection, and low labor costs, are applied in the automation field. For mass production scenarios, AOI can address the allocation of operators, learning costs, and deficiencies in production efficiency. In this study, we adopted the You Only Look Once version 7 (YOLOv7) architecture for the AOI system to conduct sample inspections. At the same time, generative adversarial networks (GANs) were used to expand the sample database. The Ethernet communication architecture was also employed to integrate equipment such as a six-axis robotic arm, programmable logic controller (PLC), and an industrial computer. This integration gave the system intelligent functions such as automatic scheduling, production report statistics, and real-time monitoring. A graphical user interface was also designed to reduce personnel learning costs, simplify operations, and enhance equipment uptime. Ultimately, through the training of the YOLOv7 detection model, we achieved excellent detection results. The detection precision reached 91%, and the mean average precision (mAP@0.5) reached 83%. This confirms the value of using GAN technology to expand the AOI sample database, especially when there is a shortage of samples in the early stages of production. This high accuracy not only helps improve the accuracy of AOI detection but also enhances the production efficiency of the intelligent processing system.