Visionairy: artificial intelligence to improve quality

Daniel Blengino and Yannis Kolodziej met at the Institut d'Optique Graduate School (IOGS) during their education in technological entrepreneurship, in the entrepreneurship track (Filière Innovation-Entrepreneurs - FIE). "We had to carry out a project during the last two years of our education. During this project, we met a number of manufacturers who told us about their quality issues. For example, we talked to a company that wanted to carry out image analysis on mascara brushes," explains Daniel Blengino. The initial solution the two students implemented in this first factory worked well and they generated a small turnover.

In 2018, while still students, Daniel Blengino and Yannis Kolodziej founded the start-up Visionairy, specialising in visual inspection. Daniel became CEO and Yannis CTO. The following year, they toured French factories to understand the sector's difficulties and to see if there was a place for their technology in the market. "Operators inspect the parts manually in all the factories. So there are problems with precision and fatigue, and young people no longer want to do this tedious job. Furthermore, manufacturers generally check only a tiny fraction of these parts. If they do not pass the inspection, the whole batch is thrown away, when some could be recovered. There's a lot at stake when it comes to waste," adds Daniel Blengino.

Some market players are already addressing these issues, offering cameras for installation on pre-existing machines. "Too complicated to set up", according to the start-up, which helps its customers choose the correct viewing system, the right number of cameras and the optimum lighting for inspecting the parts they produce.
 

Digital parts inspection

Visionairy offers specialised artificial intelligence (AI) software for parts image analysis, that can be easily connected to industrial or off-the-shelf cameras. "Our programme examines the image and detects whether or not the element is damaged. With our centralised software, which can group several lines together, the customer can view all their parts and trace the source of the defect," says Daniel Blengino.

In 2020, Visionairy installed its first solution at Toshiba, in Dieppe, to check its electronic chips. It is also developing a partnership with CNRS and ENS Paris-Saclay to design unsupervised learning technologies. "When we deployed our unsupervised AI technology, we found that it was difficult to show the AI images of faulty components because they are so rare. Our process, which is the subject of two patents, detects anomalies in good parts. It has been used in production since 2023," explains Daniel Blengino.

In 2022, the duo raised 1.2 million euros from the NCI WaterStart investment fund, as well as from business angels and the BPI. This helped them to recruit a team of 15 people and step up their commercial operations.

The following year, Visionairy was rolled out across Europe. In France, it works with the cosmetics industry to inspect aluminium cans to an accuracy of 0.1 mm, at a rate of two parts per second. On the Plateau de Saclay, it joined the Siemens ecosystem and works with a number of large-scale industrial projects. In neighbouring Germany, Visionairy automates defect detection on complex automotive parts with its unsupervised AI technology. In Switzerland and Italy, the start-up monitors a glass defilming process.
 

Complex software

Visionairy connects its camera-based software to a learning and viewing platform. This platform feeds data back to operators in real time, or the data is organised to drive artificial intelligence. "We do all this as part of Visionairy LAB, our machine vision laboratory that we use to pre-qualify our project," explains the co-founder. Once the AI has been trained on the desired parts, it is configured and deployed on a special computer that will run the AI on the production line. "Once the image analyses have been carried out on the production line, we record the results, which are fed back to the Visionairy LAB. This helps us to understand a posteriori how potential faults arose and at what point in the line. The source of the anomaly can then be traced, so the manufacturer can better control their process and digitalise their quality," explains Daniel Blengino.

The start-up sells its software on a subscription basis, which includes services. "The cost of our solution pays for itself within a year, and is particularly advantageous if customers have a lot of lines to equip," says the start-up. Visionairy is currently expanding into the energy, glass and automotive sectors. But it is also looking to enter the highly selective pharmaceutical and aeronautics sectors. "In aeronautics, AI won't be able to replace humans, but it will enhance them and help them make decisions. It will be a support tool," explains the co-founder.
 

Improving the tool

Visionairy, which currently resides at the École polytechnique incubator on the Plateau de Saclay, continues to grow and aims to achieve sales of one million euros by 2025. To achieve this, it will continue its expansion in Europe, as well as in the United States, while working to diversify its tools. "We are going to integrate object recognition into our tool. For example, on an automotive component arriving on a conveyor, our software will automatically recognise the part and apply a quality control. Whereas today, the programme only analyses the image presented by the sensors," explains Daniel Blengino. The aim is to increase autonomy and flexibility.

Visionairy will seek to boost the performance of its AI over the next few years, notably through iterative learning; instead of learning from a conventional database, the programme will learn from the images circulating on the production line and update the parameters according to the reality on the ground. The start-up will also increase the precision of its AI. The aim is to be able to see finer elements, to detect tiny defects on large images of variable parts.

The two co-founders will soon carry out a carbon assessment of their technology, subsidised by the BPI. "We would like to do this by using a calculator to work out the impact of our solution on our customers' emissions. We know, for example, that our process saves some of our glass customers from having to throw away nearly 30 glasses a day. So they save on raw materials, the energy used by the machines and so on," concludes Daniel Blengino.