Interview Faster and more Reliable Visual Inspection for Die Casters
With computer vision and deep learning approaches, the Italian company Covision Quality wants to help the industry. The joint project with Alupress is intended to support this. We spoke with Franz Tschimben, CEO of Covision Quality, about his company's goals and challenges.
What is Covision Quality about?
Franz Tschimben: Covision Quality automates the industrial quality control process on metals through computer vision and deep learning technology. Computer vision is considered to be the most technologically mature field in modern Artificial Intelligence. It is the field that makes sense of images and videos by applying algorithms.
Covision Quality was born out of Covision Lab, a computer vision focused research and development hub and company builder with the goal of shaping industries through its applications. It was founded in 2019 by seven multinational technology and industrial companies.
What solutions do you implement for die casters? Do die casters have to meet certain challenges to be able to apply deep learning solutions at all?
Franz Tschimben: The solution can be implemented by any company, of any size, that is processing metals in some way or form. What we need are images captured during the visual inspection process. Die casters do not have to meet particular challenges or criteria in order to apply the computer vision and deep learning solution of Covision Quality.
What we have seen in the market is that there are three approaches to visual defect detection for metal processing companies and specifically for die casters. In some cases companies leverage exclusively human visual inspection, some others leverage inspection that is based on vision systems and other companies who implement a combination of both approaches.
We at Covision Quality integrate with any of the three given situations. We leverage the images captured by existing vision systems or we help the company set up a specific image capturing system.
Can you provide an example of where and how we can start an Artificial Intelligence application as a tool to see some good results?
Franz Tschimben: The best way to see some good results immediately is to have our software to do a free test run on images that are given to us by a potential client. The more images the better of course. A heatmap will then immediately give an indication of the results. Following this initial assessment of the situation, we usually set up a 2-month feasibility study in order to test reliability and scalability of our software. During this phase we get familiar with the visual inspection requirements of the client, incorporate the error catalogue and provide first results of reliability and scalability. The final step is to then scale to as many production lines as possible.
How long have you been working with Alupress and what experience have you had in this time?
Franz Tschimben: We have been working with Alupress for a few months now, with the goal to leverage our computer vision and deep learning approach to make visual inspection faster, more reliable and scalable. First results show that it is working well.
Thank you for the interview, Mr. Tschimben!