Deep Learning Convolutional Neural Networks in Autonomous Vehicle Control - JHB
Details
By Byron Louis de Villiers
Agenda
• 18:00 - Pizza and networking
• 18:30 - Intro talk by Rishal Hurbans
• 18:50 - Introduction to Neural Networks by Privolin Naidoo
• 19:05 - Convolutional Neural Networks in Autonomous Vehicle Control by Byron de Villiers
• 19:45 - Discussions, drinks, networking
Introduction to Neural Networks
A brief introduction to the history and the fundamental design of Neural Networks. The talk will cover a high level overview the internal workings of Artificial Neural Networks and how this is extended to other types, such as RNNs and CNNs.
Deep Learning Convolutional Neural Networks in Autonomous Vehicle Control
Convolutional Neural Networks are machine learning or artificial intelligence algorithms which have the ability to iteratively learn from data. These are powerful tools to solve real world problems today. I will briefly touch on the advancements made from machine learning and AI. I will briefly explain what Convolution is and how it is applied to Neural Networks. I will then discuss my Honors Dissertation and how I developed and trained my own CNN to autonomously drive an RC car around a track. Simulating a small scale real world self-navigation system. This will included the development of the vehicle control system, gathering training and test data, developing and training the CNN and deploying the trained algorithm in autonomous vehicle control. I will then end off by showing the main results and findings. Finally, i will show a short video of the vehicle successfully following a track, at a constant speed, completely unmanned.
About Byron
Byron is close to completing his final year of honours at the University of Johannesburg studying a B.Ing in Electrical and Electronic Engineering. The title of his Honors Dissertation is ‘Deep Learning Convolutional Neural Networks Deployed in Autonomous Vehicle Control’. The project’s main objective is to use Google’s TensorFlow, to train a Convoluted Neural Network algorithm to predict the approximate control for an autonomous vehicle. He is well versed in Mathematics, Physics and logical and technical problem solving. With interests in Algorithm and Software Development, Artificial Intelligence, Data Science, Control Systems and Robotics.
If you're just interested and cannot guarantee attendance, please DO NOT choose to RSVP yes.