How Camera Recordings Became a Powerful AI-based Production Line Optimization Tool
Annotation automation and gathering insight from manufacturing videos for production process chronometry and optimisation.
Like many manufacturers, our partner already had all the data needed for AI-based production process optimisation, but wasn’t using it.
Our project partner Signify had a huge collection of video recordings from industrial cameras covering a number of areas and stages of production.
Such data is an excellent input for machine learning algorithms, which can be used to detect and track objects of interest, detect defects, point out missing components, and these data can be used to determine areas for optimisation, point erroneous assembly and reduce potential waste much faster and more effectively than a human eye.
The problem was that the data was useless without a costly annotation process.
There is a significant obstacle to implementing such an AI-based solution – manually describing all these videos requires thousands of hours of work by people who understand what they are looking at, and painstakingly annotating the data for future use by the algorithm and training of the neural networks. It takes a lot of work, time and, of course, a significant budget, which affects the return on investment for innovation and the decision to even start the project.
đź‘‹ Hey You, What Are You Working On?
Share a few details about your challenge or idea, no matter how early or complex. Our expert team will reach out to offer our expertise in terms of feasibility, budget, and scope. Let’s shape a solution for you together!
So we thought: “Let AI help AI understand what’s in the picture”.
We created a tool that can take a small sample of frames from video, manually annotated by a human, and extrapolate that description to a much larger dataset.
You simply upload the videos to the tool, provide a few exemplary samples of what you are interested in, and the solution describes the remaining material to the level required by machine learning. It creates a new dataset, which can be further used to train new models that will monitor the production process effectively, can be used for automated chronometry, and to gain new insights about the production process. This information can be used to effectively indicate areas for production line optimisation.
The result is a well-fed Computer Vision algorithm that helps manufacturing specialists draw conclusions about optimisation opportunities with a high degree of accuracy and gain new, automatically produced insights into the production process.
This project is part of the AIMS5.0 initiative – a European R&D project aimed at the coordinated introduction of AI into the industrial environment and improving sustainability in the semiconductor industry.
Got a Project in Mind? Let’s Discuss It.
Contact us!
Send us an email: [email protected]
