a visual course

How Tesla
Self-Driving Works

(Autonomous Driving)

How eight cameras and a neural net turn a car into a driver — in about half an hour.

A free, visual course on how Tesla Full Self-Driving works — from a camera-only view of the road, to detecting objects, building a 3D bird's-eye-view of the world, predicting what other drivers will do, planning a path, and learning from a fleet of millions.

Start with the cameras →

Part 1 — Seeing the road

1

Eight cameras, no lidar

Tesla bets on cameras alone — eight of them — stitched into a 360° view of the road.

2

Turning pixels into objects

Neural networks scan each frame for cars, lanes, traffic lights and pedestrians.

Part 2 — Building a model of the world

3

A bird's-eye view of the world

The car fuses all eight cameras into one top-down 3D map of the space around it.

4

Predicting what others do

Every nearby car and pedestrian gets predicted paths — several, each with a probability.

Part 3 — Deciding and driving

5

Planning the path

From many possible trajectories, the planner scores each on safety, comfort and progress — then picks one.

6

Learning from the fleet

Millions of cars surface rare situations, label them and retrain the net — increasingly end-to-end, photons to steering.