Главная
Study mode:
on
1
- Intro & Overview
2
- Current Auto-Breaking system
3
- Full Self-Driving from vision only
4
- Auto-Labelling for collecting data
5
- How to get diverse data from edge-cases
6
- Neural network architecture
7
- Tesla's in-house supercomputer
8
- Owning the whole pipeline
9
- Example results from vision only
10
- Conclusion & Comments
Description:
Analyze a talk by Andrej Karpathy on Tesla's progress in self-driving technology using vision-only systems. Explore the shift from multi-sensor approaches to camera-only solutions, discussing data labeling techniques for petabytes of information, edge-case sampling methods, and real-time neural network training. Delve into Tesla's in-house supercomputer, their end-to-end pipeline ownership, and the advantages of vision-only systems over multi-sensor alternatives. Gain insights into the current auto-braking system, full self-driving capabilities, neural network architecture, and example results from the vision-only approach.

Self-Driving from Vision Only - Tesla's Self-Driving Progress by Andrej Karpathy - Talk Analysis

Yannic Kilcher
Add to list
0:00 / 0:00