Jinan has unveiled China's first industrial artificial intelligence (AI) vision inspection system and DihugeNet large AI model, independently developed by CAS Dihuge (Shandong) Technology Co Ltd. The technologies deliver high-precision inspection and full-angle deep scanning capabilities, providing inspection solutions for manufacturers in sectors such as consumer electronics, optical interconnects, new energy, and automobiles.

The industrial AI vision inspection system features three major innovations. It uses a submicron-level computational imaging system to simultaneously generate high-resolution 2D images and 3D point clouds, with a resolution equivalent to about one percent of the diameter of a human hair, enabling the identification of microscopic defects.

The system also adopts a multi-axis flexible transport platform to conduct full-angle and deep scans, eliminating blind spots when inspecting products with complex surfaces and irregular structures. In addition, it integrates a multimodal inspection model that uses a large AI model to enhance lightweight models, addressing challenges such as limited samples and the need for real-time detection in industrial scenarios.

"We have integrated the industrial AI vision inspection system with the DihugeNet large AI model," said Linghu Bin, CEO of Dihuge. The company developed a knowledge distillation approach that transfers the perception capabilities of large AI models into lightweight inspection models, reducing model training time from hours or days to mere minutes.

The technology has been applied by several listed companies in fields such as optical modules, consumer electronics, lithium batteries, and automobiles. The company plans to expand its application to more sectors, including heavy industry and pharmaceuticals.

An employee operates the AI-powered industrial inspection equipment developed by CAS Dihuge (Shandong) Technology Co Ltd. [Photo/Dazhong News]