IDENTIFICATION OF TIRE RUBBER FEASIBILITY WITH CNN RESNET-50 MODEL

Abstract

In this study, the authors designed an algorithm based on a convolutional neural network that is capable of automatically classifying tire rubber eligibility according to the appearance of the tire on the tire image. The proposed algorithm will be built through several stages as follows. In the first stage, tire image acquisition will be carried out which will be the input of the designed algorithm. Furthermore, the acquired image will be divided into two sets, namely training and testing sets. The training set contains tire images that will be used at the training stage of several convolutional neural network architectures to be able to and classify them to the appropriate level of feasibility. The training phase will be carried out in a number of epohs, and at each epoh, the cross entropy loss function value will be calculated which expresses the performance of the convolutional neural network architecture in classifying tire images.

In this study, the author has designed an algorithm based on deep learning that is capable of automatically classifying tire eligibility. The proposed algorithm has been built through several stages such as tire image acquisition, training of several CNN models, especially ResNet-50. The CNN architecture test is trained to classify tire images from the test set. In addition, the accuracy value has also been calculated which shows the percentage of the number of tire images that are successfully classified correctly to the total number of tire images in the test set, which is an accuracy of 88.31%.