Novosibirsk State Technical University has developed the national AutoVisions machine vision system, which automatically detects defects on the line, monitors compliance with safety regulations and helps enterprises reduce losses, prevent accidents and improve work efficiency.
The AutoVisions project is an inter—faculty project that brings together students with competencies in IT development, computer vision, industrial automation, and project management. The project team consists of Nikita Dyubenkov (Faculty of Business), Artyom Zhidov (Faculty of Automation and Computer Engineering), Makar Minaretsky (Faculty of Mechanics and Technology), Vladislav Sataev (Faculty of Information Technology, NSU).
The developers were faced with the task of creating an industrial machine vision system for comprehensive control of production processes. AutoVisions provides stable product quality control, monitoring of equipment and compliance with safety standards in real time. The system integrates into the existing infrastructure without stopping production.
The solution is built as a modular ecosystem and includes: N1 — an AutoVisions network amplifier in factories, Apex — a camera data center + a Web-based control panel that combines data and integrates the system with enterprise equipment, INSIGHT — a camera for defect detection and security control with an AI module.
"An example of INSIGHT's work: the camera tracks each object, and the video stream is analyzed by two neural networks. EfficientAd detects deviations from the norm using heat maps, YoloV11 classifies defects as "normal", "dent", "crack", etc.," said Nikita Dyubenkov.
"We are adapting the system to a specific customer — there are enough photos of high-quality and damaged products to set up. The prototype showed an accuracy of 94.87% according to the AP50–95 metric. A website with a landing page and an online store has been developed, and a platform for centralized management of the entire ecosystem is being created," said Artem Zhidov.
The team has successfully passed the NSTU Reactor Pro acceleration and is applying for the Student Startup and A:START grants. According to the developers, two industrial enterprises have expressed their willingness to pilot the system.