Project Date: November 2019
Collaborators: Prince Steven Annor, Sampanna Bhattarai
Description of Project: The goal of this project was to implement a convolutional neural network for classifying road signs in the United Arab Emirates. We supplemented existing traffic sign datasets with new images collected from the Abu Dhabi downtown area. A data augmentation library was used to simulate varying environmental conditions, such as rain, snow, low lighting, and sun glare. This allowed the model to generalize across varying environmental conditions. After optimizing the model, the project’s goal was successfully achieved with a final classifier result of 94% training accuracy, 82% validation accuracy and 92% test accuracy.
Technologies Used: Convolutional Neural Network, Python (scikitlearn), TensorFlow, Keras