World Aerosol talks about us : Visionairy, winner of the 2024 FEA Aerosol Startup Award

World Aerosol talks about us : Visionairy, winner of the 2024 FEA Aerosol Startup Award
Visionairy, winner of the 2024 FEA Aerosol Startup Award, has developed software that is powered by a unique patented AI and plug & play on standard cameras system. Co-founder and CEO Daniel Blengino explains how this enables the automation of visual quality inspections in factories

ADF 2024 Startup booth sponsored by the FEA

Visionairy was founded in 2018, and is based in southwestern Paris. Blengino said that, after visiting several factories in France, the founders discovered that a lot of quality control was being conducted manually. Existant solutions in the market were not performing at an optimal level.

Aerosol manufacturing is very mature regarding the use of computer vision for quality inspection, he added. The main actors have been using smart cameras for many years, and even if traditional vision has provided opportunities to improve quality, these cameras also limit existing systems. For example, the camera set up is dependent on expert handling, and they require a lot of programming and can bring poor performance.

“There is a high false positive rate that leads to unnecessary scrap, which is one of the biggest computer vision challenges in the aerosol market,” added Blengino.

Aerosols can present various types of defects, even with most standard vision systems – and variability is inherent to aerosol production. Traditional vision detection requires adherence to pre-defined rules to identify and classify the object and image. Therefore, the more variability there is, the more difficult it is to pre-define those rules and cover all possibility cases. This limit explains a poor performance and a high false positive rate provided by such systems, said Blengino.

“AI-based computer vision is a new way of optimising computer vision in this market, and can help teams to overcome these different challenges.”

“That’s why we created Visionairy – with the goal of making AI computer vision accessible to all factories in the world, in order to replace the operator on higher value-added tasks, avoiding manual control, avoiding scraps in the factories and bringing agility and performance and quality management to the facilities,” he continued.

GLAD software
Visionairy platform

Visionairy’s software is a no-code digital platform that can create, test and deploy AI-based computer vision applications to automate defect detection, process control or part sorting.

The company’s patented anomaly detection AI is called GLAD, and it trains itself using only OK images, instead of learning from a defect database. Consequently, the software enables a rethinking of the entire detection process, and represents super-digitisation of Industry 4.0 manufacturing strategies.

“Our innovation is not limited simply to implementing AI, but actively contributes to the digitalisation of the operation by making the entire process autonomous and traceable,” said Blengino. “Where other industrial vision systems only provide a statue based on the image, our system goes further by collecting all the results to enable the centralisation of data on a single platform. “This allows you to access production data in real time, enhance traceability by accessing your entire control history and continuously improve the performance of your vision application.”

Optimal images

For Blengino, a good computer vision solution relies on three things: good technology, a good tool and good images.

“You can have the best image analysis technology, but if the information to detect is not in the image, it will not be able to detect anything. A good image is only a question of optical set up. [...] Before talking about AI, being able to take a good, replicable image, is key to the success of any vision project.”

As aerosols are mostly made of aluminium, or other reflective materials, it can be difficult to capture strong images. An optical engineer by background, Blengino said that Visionairy’s software-as-a-service includes access to the company’s optical expertise.

“As vision experts, we are able to recommend the best optical setup – camera, lighting and configuration – for each use case.”

Furthermore, as a software provider, Visionairy will always recommend the most cost-effective hardware using a standard camera, lighting and programmable logic controller (PLC). “We can do that because our software is compatible with any camera. It also means we can retrofit existing cameras, set up and production lines.”

Data input and process
Visionairy plateform

GLAD only needs 50-to-100 OK images to create a model that can be deployed in Visionairy wins the FEA Aerosol Startup Award production, which is not a lot of data at all.

If a factory is working well, the defects will not be easily accessible. GLAD brings the benefit of enabling some automated inspection with accessible images. Another benefit of not requiring vast amounts of data is that Visionairy can swiftly deploy in production, while detecting anomalies and new defects.

Once the AI model is evaluated, Visionairy is able to send an OK and not-OK statue to the PLC. If defects are found, the line can be stopped, or the product can be refused.

“It’s completely customisable,” said Blengino. “We are able to reject the product, or just push an alarm on the machine. We can do what we want, with the help of the automation team in the factory or our partner integrators.”

“We have over five years’ experience in the field, we are now able to provide this expertise very quickly, and are working on the implementation of an optical assistance tool integrated into our product to enable each of our customers to be autonomous in designing the configuration that serves them.”

FEA Startup Award
Visionairy wins the FEA Aerosol Startup Award

At this year’s ADF, part of Paris Packaging Week, Visionairy was named winner of the FEA’s Startup Award 2024. World Aerosols asked what impressed the jury in particular.

“We are very proud of this award, and we want to thank the FEA,” said Blengino. “We are addressing a key issue in the industry, which is quality. This industry is already very mature in computer vision, so succeeding in bringing a true innovation that pushed performance boundaries really interested the jury, while talking about reducing scraps, costs and enhancing production reliability.”

“Our ability to test, integrate and launch our solution into production with an initial industrial buffer in less than six months surely impressed the jury,” added Blengino.

The fact that the company is helping to improve the quality and address a production problem rather than a design one also resonated with the judges, he said.

Other applications

GLAD can be used in other industries – such as electronic, automotive, cosmetics – and can be applied to all metal and packaging materials in the use cases of cosmetic control, assembly control and part sorting.

The software is a great match for the aerosol industry, though, because cameras are already installed, and Visionairy can retrofit existing systems. It is also an industry that has experienced the limits of traditional computer vision. Developing a new technology that has transgressed boundaries is why the company really loves the aerosol sector.

Having said that, Blengino noted that Visionairy’s activity in other industrial segments can only benefit its work in the aerosol space, and vice versa. “In the aerosol industry, we have the experience of deploying an AI solution and production with our patented AI, so it’s a great advantage for deploying and spreading this technology in other factories in other sectors.”

As for the rest of this year, Blengino said the company is committed to growing its customer base in its key markets of aerosols and packaging. It will also be present at other trade exhibitions, including Hannover Messe, which takes place from 22-26 April, and where Visionairy will have a booth in the startup area.

Go to :

Published on
Written by
World Aerosols : the voice of the aerosol industry

Latest articles

See more