Pengaruh Perbandingan Ukuran Gambar Menggunakan Metode Deep Learning Model CNN Yolov4 pada Pendeteksian Mobil untuk Tempat Parkir

Abstract

This research is motivated because there are still many car parking lots that use human resources. Therefore the process of vehicles entering the parking lot is still less effective. Because of the problems that sometimes occur, such as when the guard is not alert at his post or when the guard is daydreaming. This can cause motorists who want to enter the parking lot to have to wait and possibly cause long queues and congestion. This research uses deep learning method using Convolutional Neural Network You Only Look Once (CNN YOLO) model. YOLO is one of the deep learning models that can be used for object recognition. This research consists of several stages, namely tool making, data collection and testing. The tool used as a car sensor in this research is ESP32-CAM. ESP32-CAM is used to capture the image and then processed by the computer to determine whether the captured image is a car or not a car. This research uses real cars with different color, angle and distance parameters. From this study it is known that the value of the F1 score (average comparison of precision and recall) at 250x150 pixels is 99.59%, 225x125 pixels is 99.59% and 200x100 pixels is 98.58%.