TrafficNN: CNN-Based Road Traffic Conditions Classification

Abstract

Shipu, Monjurur Kader Mamun, Faisal Al Razu, Shamim Hossen Nishat Sultana, Ms.Traffic congestion can affect different socioeconomic aspects of a country. It is necessary to develop an automated system that can perform better than traditional human effort in traffic controlling. So, detection of traffic conditions is crucial for building a smart traffic control system to prevent traffic jam escalation. Modern deep learning approaches can help in detection of road traffic conditions especially convolutional neural network (CNN). We suggest a novel model in this paper that can classify road traffic conditions using CNN. Our proposed model ‘TrafficNN’ classifies five different road traffic conditions with an accuracy of 82%. To train and test our model, we use the traffic conditions images which are collected from the Internet. To prove the strength of our model, we compare it with several pre-trained models like VGG16, ResNet50, InceptionV3 and DenseNet121. The comparison result proves the significance of our model to extend its successful application for developing an automated traffic controlling system in the near future.