Transforming RMG printing with computer vision: achieving unmatched precision in fabric defect detection

Abstract

This project investigates the application of advanced computer vision methods for detecting and sorting printing blights in the ready- made garments(RMG) sector, with a particular emphasis defect detection. The primary goal is to reduce the time and expense involved in manual checking by achieving the level of perfection in quality control procedures. This study evaluates recent advancements in computer vision operations for dis gurement discovery, relating the limitations and challenges in current methodologies. Our research leverages state- of- the-art im- age recogni- tion algorithms, including convolutional neural networks (CNN) and deep learning models like Simple CNN, MobileNet and E cientNet to accurately identify printing defects. The primary areas of focus include the development of a comprehensive dataset of publishing blights, the creation of robust image process- ing channels, and the integration of these systems into existing RMG product lines. Additionally, the signi cance of this exploration lies in its implicit to transform the quality control process within the garments industry. By automating dis gurement discovery, we aim to markedly reduce the reliance on human or manual examination, therefore adding product e ectiveness and lowering costs. Moreover, this advancement will not only improve the overall quality of the products but also provide a competitive advantage to manufacturers in the global demand. Our nd- ings show how we can achieve advanced discovery rates and accuracy compared to conventional styles, in this manner streamlining the quality control process. Models like Simple CNN, MobileNet and E cientNet using epoch value 10 provided maximum levels of ac- curacy, scoring 85.8%, 97.6% and 98.47% respectively. Our future exploration will concentrate on re ning these algorithms and examining their scalability and rigidity to colorful types of publishing blights and garments.