Get rolling with ML on AWS DeepRacer
summary
AWS DeepRacer is a 1/18th scale autonomous car that trains via a 3D simulator using AWS SageMaker for training, RoboMaker for the environment and S3/Kinesis/CloudWatch for storage and video. Training uses a Python reward function as Lambda that assigns rewards from simulator variables; the car captures ~50 images/sec and typically converges after ~20 episodes. The Evo adds a stereo camera and lighter sensor and a garage lets you lock an action space before training. Accenture ran a private league training 2,000 employees.
AI-written summary · about
chapters (17)
- 0:00Introduction and sponsors
- 1:08Meet the speaker and DeepRacer
- 2:15What is AWS DeepRacer
- 3:01Reinforcement learning basics
- 6:24DeepRacer League 2021
- 7:30How to win with DeepRacer
- 9:19Types of machine learning
- 11:49Reinforcement learning explained
- 14:21Reward function in action
- 17:31Training process and model
- 19:17Neural network architecture
- 21:06DeepRacer console and services
- 22:57Customizing reward functions
- 24:04Track components and variables
- 25:51Garage and sensor customization
- 26:53Accenture’s role in DeepRacer
- 27:55Resources and next steps
related talks
27mGet rolling with machine learning on AWS DeepRacer
23mGet rolling with machine learning on AWS DeepRacer
24mGet Rolling on Machine Learning
37mAWS DeepRacer: Train, Evaluate, and Tune
30mStart your ML journey with AWS DeepRacer for public sector
37mAccelerate your team’s machine learning journey with DeepRacer
28mShift your ML model into overdrive with AWS DeepRacer analysis tools
37mAWS DeepRacer:train, Evaluate, and Tune Your Reinforcement Learning Model