Defect Classification System for Ski Goggle Lens

Abstract

In this work, we build the defect classification system for ski goggle lenses with machine learning. In the first step, we establish the image capturing model and data annotation. In the next step, we apply the classification model based on machine learning for classifying defects such as scratch, watermark, spotlight, border, stain, dust-line, and dust-spot. Besides, to increase the performance, the augmentation method is also applied in the training process. The classification rate achieves 94.34%, while the running time is short.