Tolga Talks Tech is a weekly video series in which Onica’s CTO Tolga Tarhan tackles technical topics related to AWS and cloud computing. This week, Tolga discusses the AWS DeepRacer program with Ryan Ries, Senior Director – Data Science & Engineering at Onica. For more videos in this series, click here.
Ryan let’s talk about DeepRacer. What is DeepRacer?
Right, so DeepRacer is a new service that AWS has put out that you can find on the console. It is a way to introduce people to reinforcement learning. They basically have taken the camera from the DeepLens project and put it onto a small vehicle that is autonomous and will run around on racetracks. So you use reinforcement learning and train your model based on a set of tracks that are inside the DeepRacer package – so it is a console application. And then you can go about training it and seeing how well it will run on a track. And there are multiple tracks inside the console that they use for training.
AWS DeepRacer’s reinforcement learning (RL) model is a unique and advanced machine learning technique that allows for learning very complex behaviours without specifically labelled training data. It is also capable of making short term decisions while augmenting them for long term goals. AWS DeepRacer integrates with Amazon SageMaker to enable reinforcement learning, as well as utilizes AWS RoboMaker, Amazon Kinesis Video Streams, Amazon S3 and Amazon CloudWatch.
AWS encourages developers to train, evaluate and tune RL models and deploy them onto AWS DeepRacer for a real world, autonomous experience. The AWS DeepRacer League, the world’s first autonomous racing league, was created to provide a platform for developers to test their skills competitively at a global scale.
Learn more about how industries are innovating with artificial intelligence, machine learning and extended reality.