Conference material: "Proceedings of the International Conference on Computer Graphics and Vision “Graphicon” (19-21 September 2023, Moscow)"
Authors:Skripkina D.V., Levitin A.V.
Comparative Analysis of One-class and Two-class Support Vector Machines for Detecting Textural Anomalies in Leather Images
Abstract:
The article presents a comparative analysis of the one-class (one-class SVM) and two-class (twoclass SVM) support vector machine for automating the detection of defects in skin images caused by its linear deformation. The task of detecting textural anomalies of the skin is very relevant, and is used in all industries related to the processing of leather into leather products. Real skin images were used in the research. Abnormal textures were obtained as a result of linear computer deformation (stretching and compression) of the original samples. Local binary patterns (LBP) are used as texture features. The evaluation of the quality of work is carried out using the proportion of correct answers of the algorithm. The influence of the deformation depth of anomalous samples on the quality of one-class and two-class methods in the absence of interference and in the presence of salt-pepper interference is analyzed.
Keywords:
Machine learning, one-class method of support vectors, two-class method of support vectors, technical vision,template
