Inventec's AI R&D center, set up in 2018, has developed AI-based computer vision solutions for testing pre-installed notebook software, double-checking AOI (automatic optiacl inspection)-detected product defects, and detecting scratches on surfaces of notebook chassis, according to the ODM.
Inventec used to have inspectors manually test software preinstalled in notebooks to check whether the software is in normal operation. However, such manual testing carried high labor cost, requiring hundreds of inspectors for one million notebooks shipped per month.
Inventec has replaced manual software inspection, including booting, key-pressing and screen view checks, with robotic arms and industrial cameras undertaking automatic inspection based on image recognition matched with OCR (optical character recognition).
AOI parameters are usually set at high levels to reach high yield rates, often making misjudgment. Therefore, AOI must be followed by human double checks. As AOI can recognize images of measurable defects but are quite difficult to define the types of defects, Inventec has used AI to recognize defect types in place of double-checking based on human eyes.
Since there are different kinds and sizes of scratches or other defects on surfaces of notebook chassis, they are difficult to be completely detected using human eyes. Inventec has developed two AI models for detecting such scratches through using a large number of images in deep learning. One AI model is able to recognize the type, size, location and number of scratches, while the other allows adjustment in detection parameters depending on clients' levels of quality inspection.