Nexar now 'sees' 130 million road miles per month to train autonomous vehicles
Go to
Article
NEXAR FOR AUTONOMOUS VEHICLES
Train your models with the largest crowd-sourced dataset of HD images and videos of streets and roads
Request a demoLeverage Nexar’s images and videos, such as near misses, other edge cases and change detection or the best training outcomes.
The world as it is, ready for training
Nexar collects data from 130 Million Miles monthly, with a dataset of 3.2 trillion images. This data is is invaluable data for AV training. Our network creates a constant flow of context rich images and video, with added sensor and geo-spatial data, from images of collisions, near collisions, busy intersections and the world around us.
Improve Your
Model’s Performance
All Nexar images and videos are automatically localized on the map and filtered for quality. Additional sensor data is also made available with the right time stamps and additional context and metadata.
Nexar dash cams “see” them all
- From near-misses and low-impact collisions (fender bender) to high impact collisions (car crashes)
- Provide valuable visual data edge cases such as collisions, including positional information of drivers, hard brakes, lane drift etc.
- Identify real-world changes that affect AVs, such as construction zones, road sign changes, road blockage, lanes and more.
- Includes diverse data across hours, days and months, weather conditions and seasons, geo-location and additional variables.
Unique
Visual Data
Dash cam users contribute data, images and video of the world
Fresh Data
and Images
Frequent use ensures a constant stream of up to date vision
Competitive
Advantage
Our unique data can be just what you need to stand out from the competition
Human
Scale
We see the world just like humans do, not like a traffic camera
Move
through time
Our constant vision stream lets you check what happened in the past
Respectful
of users
Our data is always anonymized and compliant, so the source of an image, license plates, faces, etc. cannot be identified
