Garment Defect Detection Using Convolutional Neural Network

The apparel industry plays a significant role in Sri Lanka's economy, contributing over 52% of total export earnings. Its primary focus is on producing high-quality finished garment products within specified timeframes. For high-quality garment products, the defects on the garment products should be at a very low level. In this paper, the automatic garment defects detection method is introduced using the principles of image processing with the CNN algorithm to identify the garment defects. Here, the authors have selected the training center of a MAS Linea Aqua apparel company which is among the major apparel companies in Sri Lanka, and developed this research to identify the garment defects done by the trainee sewing operators during their training period inside the company. The defects that occurred during the use of the flat-seam machine are considered and relevant defects are divided into several categories including wavy-seam defect, up-down defect, open defect, and non-defected. Also, researchers have introduced a Machine learning algorithm to calculate the defect rate done by a certain machine operator during their training period to analyze their sewing skills.