Implementasi SSD Mobile-Net untuk Mendeteksi dan Pengklasifikasian Kerusakan Jalan secara Real Time Berbasis Android

Abstract

The road is a means of transportation on land, where people, vehicles, etc. Roads are also a means of supporting all community activities, both economic and social. One of the elements that can cause road damage is the large amount of traffic using the road. Identifying road damage on a road is the initial stage in road repair so you can decide what steps need to be taken. the type of pothole, experts still use their sense of sight to classify road damage in their daily work and have not automatically determined the type of road damage. As a result, if a system can detect and classify road damage it can make it easier as the first step in assessing pothole damage. The architecture used to build the system is the MobilenetV2 architecture, which uses a mobile-based Single Shot Detector (SSD) method in real time. In this method the data used is 2400 images of four types of pothole damage, the four types of road damage used are large-shallow, small-shallow, large-deep, small-deep. By using the SSD-mobilenet steps, the results of this study were able to obtain an accuracy value of 96.08%. With the results of this accuracy, the system built is very good at detecting damage to potholes.