I’d like to share this university project made in collaboration with NVIDIA Brazil, to show how powerful and friendly the Jetson Nano can really be.
First of all, this was a group project done at the Insper (a Brazilian university) computer engineering course, where the goal was to develop an autonomous vehicle capable of understanding Brazilian road sign regulations. Along side with the main objective, we were told to use every NVIDIA toolkit we could, in order to present a nice demonstration, not only for our mentors from NVIDIA Brazil, but also for our colleagues and enthusiasts who weren’t familiar with these technologies before. Since we had zero experience either, the nice thing about that was the ability to use the Jetson platform to learn and deploy some of the NVIDIA AI and Hardware Acceleration toolkits, like DeepStream, TAO, and TensorRT, with last but not least the Jetbot repository, which helped us a lot since most of the hardware configuration could be done following its steps.
By combining all of that, we managed to achieve a self-driving Jetbot, which used DeepStream to recognize the Brazilian road signs (following a custom TAO model) and a PD controller for the road navigation.
Here’s a picture showing how the project looks:
If you have any kind of question related to this project, please feel free to post a reply or leave an issue at our Github repository.
Thank you for taking the time to read this post.
Edgard Ortiz Neto