Quality assurance is a crucial process in modern manufacturing, with defect detection playing an important role. One effective method for defect detection is the use of deep learning models that analyze product characteristics based on a large number of images. Due to the rarity of defects, there is a problem of insufficient examples of defective products. To address this issue, we employed a deep learning model trained exclusively on images of good products, for which a large dataset is available. We developed an autoencoder that was trained on images of good products, and then analyzed and tested the performance of this model in detecting defects based on images.
|
