KLASIFIKASI DAN DETEKSI KERETAKAN PADA TROTOAR MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK

Classification and Detection of Cracks on Sidewalks Using the Convolutional Neural Network Method

Abstract

The sidewalk is a part of the highway that is specifically provided for pedestrians where the sidewalk is generally located in the road benefit area to make it easier to walk. This is so that pedestrians do not mix with vehicles which of course can slow down the flow of traffic and can endanger the pedestrians themselves. But in reality the pavement surface has a variety of conditions. Therefore, sidewalk repair is the right solution so that sidewalk damage does not get worse and does not disturb sidewalk users. The first step in pavement repair is detecting damage that is on the surface. One of the methods that can be used to detect pavement damage is to use the latest technology, one of which is deep learning using the CNN method. The purpose of this study is to develop an algorithm that is specifically used to distinguish cracked and non-cracked sidewalks. The training dataset used is 3200 images and 800 images for test data. Where did we take this image from the Kaggle catalog. From the research we conducted, the test results showed that the model succeeded in distinguishing cracked and non-cracked sidewalk surfaces with fairly high accuracy, where the average accuracy value was above 96% and the loss value was close to 1.5%.