Assessing Deep Learning Model Using AlexNet for Water Traffic Counting in Martapura River
Nahdi Saubari, Wang Kunfeng

College of Information Science and Technology, Program of Controlling Science and Engineering, Beijing University of Chemical Technology

Abstract

In recent years, the traffic of water transportation in Martapura river has been increased and creating many problems for the city and its environment. Hence, the traffic needs to be managed from time to time. Deep learning model might be used for traffic counting by detecting the ships. This study aims to assess AlexNet for traffic counting purposes in Martapura river. Data were collected two times a day for 3 months by using smartphone camera. Series of experiments were developed using Alexnet model to classify and detect ships or boats in Martapura River to draw a baseline for water traffic counting system. Result shows that Alexnet gives around 97% accurateness in detecting ships or other water vehicle as the main transportation. This certainly helps the traffic counting in Martapura river. Around 5 to 7 water vehicles were detected per hour. AlexNet also detect other floating objects like water plantation or plastic garbage. Other than object detection, AlexNet as Deep Learning technology can be used for water traffic counting globally.

Keywords: Deep learning- traffic counting- river- AlexNet

Topic: Engineering