Contact Lens Automatic Optical Inspection Method Based on YOLOv9

Traditional inspection methods for contact lenses primarily involved manual examination, where workers directly assessed the physical properties and manufacturing defects of the lenses. Consequently, the quality determination relied on the workers' judgment, leading to inconsistencies. In order to address this issue, although Automated Optical Inspection (AOI) was introduced, it has limitations in applying to various types of contact lenses. As part of research to utilize deep learning in AOI, this paper introduces a You Only Look Once (YOLO) v9-based contact lens inspection technique that leverages a small dataset of contact lens images. This method involves training YOLO v9 with the limited number of images that include various defects such as bubbles, cracks, and foreign particles, as well as images of defect-free contact lenses.