People Counting for Train Temperature Control using You Only Look Once v5 Method

Abstract

AC (Air Conditioner) is an air conditioning system that is used anywhere, one of them on the train. Air conditioning is also a major factor in the comfort level of train passengers. However, train air conditioning currently can only be controlled manually from the driver's cabin so that the number of passengers does not affect the air conditioning temperature. What's more, the air conditioner absorbs 30% of the total power in the train. Therefore, this research was compiled with the aim of being able to calculate the number of passengers through the concept of people counting using the You Only Look Once (YOLO) method for the train's temperature control system. This research was built using 1 object class, namely person and the YOLOv5 algorithm, where this method is an object detection method with a high degree of accuracy. This research was designed using the Python programming language with the Google Colaboratory platform and MQTT for the data storage protocol for passenger detection results. The dataset used is passengers in the train as many as 1300 images. The training and validation results model obtained a 99.63% mAP, 95% precision, 100% recall, 97.5% F1 score, and 79.96% average IoU. Then, the model was tested and obtained 99.35% mAP, 95% precision, 99% recall, and 97% F1 score. This research resulted in a fairly good system performance in detecting and counting passengers on trains.