AI and Self-Driving Cars
summary
NVIDIA argues deep learning replaces hand‑coded perception by training models on real and simulated data and deploying them on an in‑car supercomputer. A single‑camera end‑to‑end net called PilotNet maps images directly to steering and drives a driverless demo through a varied lot and down Lombard Street; stacks of nets run on Drive PX with software in DriveWorks and Jetson dev kits, using GPUs, simulation (even Grand Theft Auto) and fleet data to cut accidents (≈40,000 US deaths, 1.2M worldwide).
AI-written summary · about
chapters (14)
- 0:00Introduction to AI and self-driving
- 2:29Deep learning fundamentals
- 6:39Driving as an AI challenge
- 10:29How neural nets see images
- 13:15Deep learning vs hard-coded algorithms
- 15:22AI Co-Pilot inside the car
- 18:32Neural nets for free space detection
- 20:20PilotNet demo in Las Vegas
- 23:29Driving Lombard Street with AI
- 25:47Centralized supercomputer in cars
- 27:11Levels of autonomous driving
- 27:53Simulation and Grand Theft Auto
- 28:15Roborace and AI in motorsport
- 29:39Q&A session
related talks
28mMachine Learning Inside the Car
52mPhysical AI: Crafting Resilient Systems with Emotional Intelligence
19mAI Trends in Intelligent Transport Systems
40mFueling the AI Revolution with Gaming
20mAccelerating AI pipelines w/ NVIDIA on AWS: An autonomous driving case
31mDeep learning for self-driving vehicle technology
02mAI-defined vehicles and autonomy
45mArtificial intelligence: How data is evolving the future of technology