Introducing ‘AIM-T1’ — Same-Day Deployment Without Labeling, Detecting Defects as Small as 3μm

A.I.MATICS, an AI specialist and subsidiary of Dreamtech, announced on the 15th that it will unveil its deep learning–based vision inspection system, AIM-T1.

AIM-T1 is powered by aimNet, A.I.MATICS’ 5th-generation deep learning engine, refined over 23 years in autonomous driving applications since the company’s founding as an in-house venture of Hyundai Motor Group in 2003. Traditional rule-based inspection systems, which rely on predefined parameters, are highly sensitive to environmental changes and often require prolonged production downtime for model updates. In contrast, AIM-T1 overcomes these limitations with an all-in-one, on-premise architecture that processes everything—from object recognition and decision-making to production line control—within 0.1 seconds, without relying on external servers.

Its detection performance is also noteworthy. The system reliably identifies microscopic contaminants and hairline scratches as small as 3 micrometers—approximately one-thirtieth the thickness of a human hair—and classifies defect severity into more than ten levels. In precision component lines at global IT manufacturers, AIM-T1 achieved a 100% detection rate for irregular defects, while improving productivity by 26% in connector assembly lines. Through its Automatic Training System (AATS), model retraining time has been significantly reduced from 72 hours to just 4 hours.

A.I.MATICS officially introduced AIM-T1 to the public for the first time at the 2026 Korea Electronics Manufacturing Industry Exhibition, held at COEX in Seoul from April 8 to 10, and has begun full-scale commercialization. At the exhibition, the company demonstrated its capability to simultaneously recognize seven inspection categories and perform real-time process control, drawing strong interest from industry professionals.

CEO Lee Hoon stated,
“AIM-T1 represents the culmination of 23 years of Vision AI innovation, optimized specifically for manufacturing environments. We aim to set a new standard for intelligent inspection lines that can be deployed and operated immediately on-site—without additional costs or specialized personnel.”

The system can be deployed instantly in the field without separate data labeling, and can begin operation the same day using only a small number of sample images. It is designed as a self-sufficient, on-site solution, enabling quality control personnel to perform labeling and process optimization independently, without external experts. This also reduces both the additional costs and delays typically associated with model changes.

Source: Economic Review (https://www.econovill.com)