Bilgisayar-görme algoritmaları ile kumaş hatalarının tespiti ve sınıflandırılması

Image processing has been employed in a variety of fields since the advent of image processing techniques. One of these fields is textile. The existence of any defect in a fabric is one of the most important factors affecting the quality of the fabric. There are many types of fabric defects that can occur for various reasons. It's critical to figure out what caused the defect and fix it so that it doesn't occur again. Automation of fabric defect detection has recently attracted a lot of interest in view of the development in artificial intelligence technology in order to be able to discover defects with a high degree of success and to limit the harm to the manufacturer. However, some problems are encountered in this area. Fabric defect detection is a challenging subject since there exist a great number of defects that might result from a variety of issues. Additionally, the restriction of this study is that the Tilda database is one of the limited datasets that contain fabric defect samples and can be accessed in this field. This thesis focuses on analyzing different feature extraction methods and different classifiers and discussing the advantages and disadvantages of the combinations. Different cases have been created that handle the data sets from different angles and apply different methods. While three different methods (EL, KNN, and SVM) have been tested in the classification stage, different CNN-based approaches (ResNet18, ResNet50, GoogLeNet, and AlexNet ) have been tested in the feature extraction stage. The results obtained have been also compared with the results of ResNet18, ResNet50, GoogLeNet, and AlexNet.