Designing the system for detecting unsuitable marble stones for an industrial process using convolutional neural networks

To the actual date, computer vision and neural network data analysis are actively developing and finding more and more applications in the industrial control engineering. These technologies are now used to deal with important problems that previously could not be solved without the help of a human as an operator. This study is devoted to the selection and configuration of a convolutional neural network for recognizing unsuitable marble samples on a conveyor belt, as well as creating an application for defining the position of these objects. Software for the classification of marble stones based on the YOLOv5 neural network model has been developed. The value of the precision and recall metric in the final experiment reached 0.957 and 0.947, respectively.