Traffic Congestion Classification Using Convolutional Neural Networks: A Simplified Approach to Traffic Management

With the widespread success of the automotive industry; one major issue that developing countries like Indonesia face is traffic congestion. This problem amplifies many issues in modern society including traffic accidents, air pollution, and also economic development. Traditional traffic management approaches have proven to be insufficient in addressing these challenges effectively. Therefore, this paper proposes a novel approach for traffic management using object detection with Convolutional Neural Network (CNN). The paper presents an insightful visualization of class occurrences, highlighting the distribution of congested and uncongested traffic images across train, test, and validation datasets. Through data augmentation techniques and hyperparameter tuning using a Keras tuner, the CNN model is further enhanced, resulting in improved performance. The retrained model achieves a high accuracy level of 95.92% on the test dataset, consolidating its effectiveness in traffic management. This research offers a promising solution to alleviate traffic congestion.