Automotive Vision · OEM

OEM Advanced Development

For OEM advanced development teams proving new driving features on real vehicles. The X2000 Vehicle Computer gives you cameras, AI inference and drive-by-wire on one unit, with 100% of the source code, so your engineers start from a working vehicle instead of a blank page.

DatasheetBrochureResearch paper

Deepview X2000 Vehicle Computer

Who it is for

Built for the work before production

Vehicle programs

Fit the kit to a development or mule vehicle and have cameras, compute and vehicle control running together early in the program.

Advanced development

Try a new perception model or driving feature on a real vehicle, with source code you can change and data you keep.

Research and benchmarking

Run and compare models in the vehicle in real time, and log every drive for analysis.

What is in the kit

One unit, from camera to steering

01

NVIDIA Orin AGX

Real-time AI inference on the vehicle computer, using TensorRT.

02

Up to 8 GMSL2 cameras

An AR0234 forward camera is included, with forward radar or lidar supported.

03

CAN-FD drive-by-wire

Two CAN-FD interfaces and vehicle control at 100 Hz.

04

100% source code

Every software module is yours to read, change and build on.

05

Data logging

1 TB on board, with video and CAN data synchronised to your data center over LTE.

06

In-vehicle display

A development UI in the car shows what the system sees and does.

How a program runs

From first install to fleet learning

Install

The kit is fitted to your development vehicle. Drive-by-wire has been integrated on Ford vehicles, with other manufacturers on request.

Drive

Lane centering runs in the vehicle from day one, using the adapted OpenPilot model in the reference design.

Develop

Your engineers change the source, add models and build the features your program needs.

Log

Every drive is recorded and uploaded to your data center.

Retrain

Models retrain on your fleet data in the Fleet Learning, and improved models go back to the vehicles.

Fleet Learning

Software reference designs

Available now, and on the roadmap

Available

  • Real-time camera interfacing at 20 FPS
  • Hardware-accelerated video encoding
  • Real-time AI inference using TensorRT
  • Lane centering AI model (OpenPilot)
  • Drive-by-wire at 100 Hz
  • In-vehicle display: development UI
  • CAN and video logging to the data center

Roadmap

  • Multi-camera: 8 cameras at 20 FPS
  • Multi-camera: 8-camera video encode
  • Depth processing with radar or lidar
  • Emergency braking reference design
  • Automated highway navigation
  • In-vehicle display: 3D visualization
  • Functional safety, ISO 26262
In-vehicle development display showing the lane centering path

Engineers alongside yours

Deepview resident engineers work inside your program on embedded software, AI training and AI inference on NVIDIA hardware. Engineering services

Schedule a call

Tell us about the vehicle and what you need to prove. Or write to sales@deepviewai.com.