IMPLEMENTATION OF THE YOLO MODEL FOR CROWD DENSITY ESTIMATION (CASE STUDY: RAILWAY STATION)

• Main Author: Aminudin Muhammad, Shafwan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/54352
• Institution: Institut Teknologi Bandung

It is difficult to estimate the level of crowd density in busy and moving crowds at railway stations and conventional method and may result in inaccurate results. This final project aims to implement the You Only Look Once (YOLO) model to estimate crowd density and improve the performance of the YOLO model in estimating crowd density. Hyperparameter tuning method is used by using the Tree-Structured Parzen Estimator (TPE) algorithm in the Optuna framework to optimize the model. Hyperparameter tuning method produces the best hyperparameter settings for the YOLO model on the IoU training threshold hyperparameter, initial learning rate, final learning rate, SGD momentum and weight decay with values respectively 0.7, 0.001, 0.6, 0.98 and 0.000123. Based on this final project, it is known that the the YOLO model was successfully implemented in a crowd management system that focuses on estimating the level of crowd density at railway stations without experiencing overfitting and hyperparameter tuning provide better performance than the initial YOLO model. There is an increase in model performance after optimization with an increase in mAP@0.5 and mAP@0.5:0.95 metrics, respectively, 0.779 and 0.0675.