Image-based vehicle traffic monitoring

Francisco Benita,Srinivasan Padmapriya, Geovani Benita and Costas Courcoubetis

Singapore University of Technology and Design

Summary

In this tutorial, we describe how to monitor traffic conditions in Singapore by retrieving images from speed cameras located at major traffic junctions. Our, module consist of three main stages. First, it connects to a public API managed by Land Transport Authority (LTA) to get stored images, timestamps and camera id. Second, it implements a Region-based Convolutional Neural Network (R-CNN) for object detection. More precisely, it uses a Mask-RCNN model TensorFlow sessions on the cloud (e.g., Colab/TPU). Lastly, we provide some samples of how to visualize the traffic data.