What Deep Learning SVI Enhanced Defect Detection Is
Machine vision systems were supposed to automate defect detection – but instead struggle in many challenging situations where the boundaries between true defects and pseudo-defects are blurred.
These struggles happen because traditional machine vision defect decisioning is based on a series of binary, yes/no choices – essentially forcing the system into guessing when it comes to difficult defects, which in turn leads to high false reject rates and costly human reinspection.
Why human reinspection? Because human quality experts are good at being able to distinguish between true defects and pseudo-defects. And our Deep Learning AI is trained just like your quality experts are, with images labeled by your quality experts, and it then applies that training to your machine vision system -- meaning it performs as well as your best inspector on their best day. It lowers pseudo-defects, removes the need for human reinspection, and dramatically reduces total cost of quality. And because SVI works with any existing camera system, there’s also no costly rip-and-replace required.
Real-World Results that Eliminated False Rejects and Pseudo-Defects
Global glass manufacturer and fabricator Vitro was having a problem with their machine vision defect detection: It struggled to accurately tell the difference between water droplets and edge chips on auto glass.
We trained SVI to learn the difference – and even better, were also able to train SVI to detect defects called rolled edges, which their existing system could not do. Accuracy increased so much that Vitro was able to remove human inspectors from their lines – and with savings in the millions of dollars annually, are currently expanding SVI implementations to more lines in more of their factories globally. In fact, Vitro believes so strongly in SVI that on top of being a customer, they also decided to participate as investors in Mariner’s Series A funding round.
sage automotive interiors
Sage Automotive also had a problem with false positives and pseudo-defects with their machine vision systems.
The system struggled to accurately tell the difference between stains and lint on the fabric, which again led to the need for costly human reinspection and line speeds that were much lower than full speed. After the Deep Learning model was trained and SVI put into place on their lines, system accuracy increased dramatically, human inspectors were moved to other high-value assets, and line speeds nearly doubled. Millions of dollars are saved annually because of Spyglass Visual Inspection – and after seeing the results, Sage Automotive’s former parent company, Milliken, has put SVI into place on multiple production lines, as well.
Can SVI Improve Your Defect Detection and Eliminate Your Pseudo-defects?
Spyglass Visual Inspection is not limited to fabric and glass use cases. Does your organization currently use a machine vision defect detection system, and can your Quality experts tell defects from pseudo-defects in the images from that system? If so, SVI’s Deep Learning AI can also be trained to tell the difference between your defects and pseudo-defects, no matter the product – and again, no rip and replace of your existing machine vision system is required.
But you don’t have to take our word for it. If you answered yes to the above questions, you’re a good candidate for our no-charge 30-Day Proof of Value; we’ll take your labeled images, create a Deep Learning AI model with those images, and show you upfront the improvements that you will see with Spyglass Visual Inspection.