Intelligent System for Monitoring the Availability of Bus Passenger Seats using the YOLO Method

The existence of public transportation is very important to support the increasing mobility of the community because it has a large carrying capacity, one of which is the bus. However, people’s interest in using public transport began to decline along with the increase in private vehicles. One of them, due to inadequate facilities, especially in terms of information on the availability of passenger seats on the bus. In this paper, a system for monitoring the availability of passenger seats on buses is proposed to support a program from the Indonesian Ministry of Transportation, namely the National Movement Back to Public Transport to improve mass transportation performance. This system detects and calculates objects using the You Only Look Once version 5 (YOLOv5) method. The dataset used in this study has only one class, namely “Person”. The results of the model training test obtained the best model with a scenario of batch size 16 and epoch 200, which produced a precision value of 0.883, a recall of 0.815, a mAP of 0.905, and an F1 score of 0.85. While the model accuracy value when integrated with NVIDIA Jetson Nano is 81.56%. The system is integrated with WebGIS to display the calculation results for the number of passengers, available seats, and bus locations.