Sustainable Deep Learning for Self-Driving Cars
Dr Duc-Son (Sonny) Pham discusses the impact of high-complexity machine learning for driverless cars and how to make this more sustainable.
At the heart of modern self-driving cars is a complex artificial intelligence system consisting of deep neural networks that perform billions of computations per second to process input images and other sensory data to help the vehicle understand the surrounding environment and make the right control decision. Many state-of-the-art deep neural networks can easily outperform human in many recognition tasks. However, their high performance usually comes at a cost: prohibitive computational complexity. This will translate to expensive hardware and high energy consumption which is not environmentally sustainable. Current research is focusing on developing new architectures that achieve similar level of performance whilst having much less computational cost and model complexity. In this seminar, Dr Pham will explain several important deep learning technologies that underpin successful designs to address the challenge and illustrate how they can be used to develop advanced "green" semantic segmentation models. It is expected that the audience will be able to apply key ideas in this talk to solve their problems in computer vision and related domain in a more environmentally sustainable way.
Date and Time
Location
Hosts
Registration
- Date: 25 Nov 2021
- Time: 05:30 PM to 07:00 PM
- All times are (UTC+08:00) Perth
-
Add Event to Calendar
- Curtin University
- Bentley, Western Australia
- Australia
- Building: 501
- Room Number: 101
- Click here for Map
Speakers
Dr Duc-Son (Sonny0 Pham of Curtin University
Sustainable Deep Learning for Self-Driving Cars
At the heart of modern self-driving cars is a complex artificial intelligence system consisting of deep neural networks that perform billions of computations per second to process input images and other sensory data to help the vehicle understand the surrounding environment and make the right control decision. Many state-of-the-art deep neural networks can easily outperform human in many recognition tasks. However, their high performance usually comes at a cost: prohibitive computational complexity. This will translate to expensive hardware and high energy consumption which is not environmentally sustainable. Current research is focusing on developing new architectures that achieve similar level of performance whilst having much less computational cost and model complexity. In this seminar, Dr Pham will explain several important deep learning technologies that underpin successful designs to address the challenge and illustrate how they can be used to develop advanced "green" semantic segmentation models. It is expected that the audience will be able to apply key ideas in this talk to solve their problems in computer vision and related domain in a more environmentally sustainable way.
Biography:
Dr Sonny Pham is currently a Senior Lecturer with the School of Electrical Engineering, Computing, and Mathematical Sciences at Curtin University. He has over 15 years of research experience in Computer Science. His current research interests include sparse learning theory, large-scale data mining, convex optimization, and advanced deep learning with applications to computer vision and image processing. He is a Senior Member of the IEEE. He is a recipient of the Young Author Best Paper Award 2010 for a publication in IEEE Transactions on Signal Processing.