Neurala, the company behind the software as a solution platform brain builder, has come up with a new way of detecting product defects which requires less data to be input in the initial training models.
According to Neurala, the issue arose when company’s failed to successfully integrate AI based detection system, because of the high quality of the product which meant that there wasn’t enough visual training data of defective products for the AI system to learn and differentiate between that and it’s quality product.
“Vision AI offers higher accuracy and more flexibility for building or optimizing visual inspections versus just using machine vision. In the past, implementing AI-based solutions has been challenging; many manufacturers tell us that their efforts have failed. Due to the generally high quality of products, they struggle to get enough images of defective items, as those images are limited in quantity”
“Neurala’s new Anomaly Recognition capability, part of our Brain Builder software platform, allows anyone to train an inspection model using only images of “good” examples.”
What this means is that any product which has a visual deviation from a “good” sample, would automatically be highlighted as a “bad example and after the training data has detected this, the brain would essentially be instructed to perform the task programmed for defective products.
“Brain Builder dramatically reduces the time, cost and skills required to build and maintain custom industrial inspection solutions”