All products (batteries, semiconductors, PCB boards, smartphone liquid crystals, etc.) produced in the factories need to be inspected for defects before shipping. Currently, most of these inspections are conducted by people who inspect products with their own eyes to find defects, and this causes high inspection costs and has limitations on inspection performance and inspection speed. One of the automation methods to address these is the inspection method based on template matching; however, this requires expert work and the types of products and defects that can be applied are limited.
SaigeVision is a machine learning-based inspection program that can address the above shortcomings. Without the need for expertise in products or defects, anyone can inspect various products for defects by simply entering the defect information of the existing products (defect status, location, form, etc.). SaigeVision trains information about the defects and images of existing products entered by users, and, after such training is completed, once the image of a new product is entered, it automatically inspects for defects in the new product.
SaigeVision provides five modes depending on the application so that the optimal solution can be applied to various defect inspections.
- Classification
- Detection
- Segmentation
- OCR (Optical Character Recognition)
- Image Generation