русскийру

"I wanted to apply my experience in computer vision in production and see a specific effect": a participant in the Design Laboratory – about the case of the Gorky Automobile Plant

06.11.2026

For the past month, participants in the new set of the Intellect Foundation's Design Laboratory have been working on cases: young researchers are looking for answers to non-trivial industrial challenges and testing hypotheses in real-world production.

We are launching a series of publications in which we will describe in detail the key developments of the participants in the current set. This cycle opens with a project by Diana Esaulova, a 4th-year student at the Moscow State University of Mechanical Engineering: together with her team, she creates a computer vision system for the Gorky Automobile Plant. Algorithms automate visual inspection on the assembly line – they identify hidden body defects and compare the actual configuration with digital 3D models even before the car is shipped.

In an interview, Diana talks about how engineering tasks are intertwined with optics and light physics, and how participation in the laboratory allows us to convert the scientific background into measurable benefits for domestic production.

– What project are you working on and why did you choose it?

– My team and I are developing a quality control system for the Gorky Automobile Plant. She must find defects on the body and check the exterior of the finished car. I chose this project because I wanted to apply my experience in computer vision in production and see if the system could speed up inspection and help employees notice more defects. I'm also interested in checking whether 3D engineering models can be used as a reference to add new cars and parts without manually marking up photos.

– What difficulties have you already encountered in your work or may you encounter in the future?

– In the search for defects, a lot depends on how we remove the body. We will have to choose cameras, optics and lighting so that even small scratches and dents are visible in the pictures. There's a lot of physics here. When checking the configuration, it is important to accurately match the images with the 3D model. Even a slight displacement of the machine changes the position of the parts in the frame, and the system may make a mistake. It remains to be seen from which angles similar details can be better distinguished.

– What problem are you and your team solving? What results will your development achieve?

– After assembly, the quality control officer manually inspects the body and checks the details with the complete set of the car. There is a limited time for inspection, so a small scratch or a similar part from another configuration is easy to miss. If the problem is noticed after the machine is shipped, it may cost more to fix.

Our system will analyze the images from the cameras and tell the employee which area or detail should be rechecked. We hope that this will speed up the inspection and help identify more problems at the factory.

– What is your role in the team? Is the project related to the area of your research interests?

– I am responsible for checking the configuration. I'm thinking about how to combine the camera images with the 3D model in order to find the right detail in the photo and check if the right option is installed. This is my second project at the Intellect Design Laboratory. Last time we developed a video cow monitoring system, and I saw in practice how much the result depends on the shooting conditions. In parallel, I am researching the reliability of machine learning models at the Institute of System Programming of the Russian Academy of Sciences. For me, participation in the Design Laboratory is an opportunity to transfer scientific developments in the field of artificial intelligence into real business processes and see what practical benefits they can bring.