Analysis of Subsampled Image Size for Detection and Identification of Brake Pad Contours by Using Deep Learning

Abstract

This paper proposes an analysis of subsampled image size for detection and identification of brake pad contours by using deep learning for the automatic detection systems. In brake pad manufacture, some problems may occur such as expansion, missing corners at the edges so that the edges of the brake pad need to obtain before checking missing corners. The size of subsampled image for training and testing for machine learning is very significant factor for optimized and efficient feature extraction. The size of subsampled image has great impact on detection and identification for feature extraction determines the accuracy of prediction of deep learning. The images are evaluated through loss function in order to observe the training process of the models. The experimental results show the method to determine subsampled image size to have better accuracy.