Donkey Car NCS

Martin Wojtczyk

Martin Wojtczyk

Berkeley, CA

0 1
  • 0 Collaborators

Neural Compute Stick enhanced Donkey Car to participate in the DIYRoboCars challenge. The Donkey Car was developed by Will Roscoe and Adam Conway. ...learn more

Project status: Under Development

Robotics, Artificial Intelligence

Groups
DeepLearning, Movidius™ Neural Compute Group

Intel Technologies
Intel Opt ML/DL Framework, Movidius NCS

Overview / Usage

The Donkey Car (source: http://www.donkeycar.com/) is an open source autonomous model car platform developed by Will Roscoe and Adam Conway. It is using a Raspberry Pi 3 and opencv & tensorflow for computer vision and inferencing.

The scope of this project is to extend the Raspberry Pi 3 with a Neural Compute Stick for fast inferencing and to compare its performance with the standard Donkey Car platform and to participate in the DIYRoboCars challenge.

Methodology / Approach

  1. Build a standard donkey car as per current instructions (http://www.donkeycar.com/)
  2. Collect dataset, train network
  3. Benchmark predictions
  4. Power-up the raspberry pi3 with an NCS
  5. Convert and deploy network, compare to base line
Comments (1)