Tesla patents method to build 3D maps of roads using camera data
Tesla describes a system that uses cameras to identify fixed objects like poles and buildings. It combines these images over time to create a consistent top-down map of the vehicle's surroundings.
In the filing’s own words
Systems and methods for a vision-based machine learning model for aggregation of static objects and systems for autonomous driving. An example method includes obtaining images from image sensors positioned about a vehicle. Features associated with the images are determined, with the features being output based on a forward pass through a machine learning model. The features are projected into a vector space associated with a birds-eye view based on the machine learning model. The projected features are aggregated with other projected features associated with prior images. Images depicting static objects in the birds-eye view are output.
What exactly they claim (claim 1)
1. A method implemented by a vehicle processor system, the method comprising: obtaining images from a multitude of image sensors positioned about a vehicle; determining features associated with the images, wherein the features are output based on a forward pass through a machine learning model; projecting, based on the machine learning model, the features into a vector space associated with a birds-eye view; storing the projected features in a queue, wherein the queue spatially or temporally indexes information; updating the queue based on the vehicle meeting a spatial threshold, wherein the spatial threshold is dynamically adjusted based on a speed of the vehicle and a location environment of the vehicle, wherein the spatial threshold is reduced in response to the location environment when the location environment corresponds to a lower speed and increased when the location environment corresponds to a higher speed; aggregating, based on the updated queue, the projected features with other projected features associated with prior images; and outputting, based on a plurality of heads of the machine learning model using the aggregated projected features and the other projected features, images depicting static objects in the birds-eye view.
Facts
- Company
- Tesla (Tesla, Inc.)
- Official title
- Vision-based machine learning model for aggregation of static objects and systems for autonomous driving
- Patent number
- US12789834B2
- Application number
- 17820849
- Filed
- Aug 18, 2022
- Granted
- Nov 10, 2026
- Inventors
- 7
- Claims
- 19
Classified under
B60W 60/001 Operations & transportG06T 7/13 Physics & computingG06T 7/20 Physics & computingG06V 10/774 Physics & computingG06V 20/56 Physics & computingG06V 20/58 Physics & computingMore from Tesla in Cars & transport
- Tesla patents electric fan system to boost car grip and cut drag Sep 29, 2026
- Tesla patents laser system to clean car windows and solar panels Sep 3, 2026
- Tesla patents system that adjusts driving based on camera visibility Sep 17, 2026
- Tesla patents rear wing that hides in body and adjusts for speed Sep 24, 2026
- Tesla patents car doors that open when you lean in using UWB signals Oct 1, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.
