Deep Learning with MATLAB, NVIDIA Jetson, and ROS

Jon Zeosky and Sebastian Castro discuss how algorithms designed in MATLAB^® can be deployed as standalone CUDA^® code to target NVIDIA^® GPUs, and how this standalone code can be used in a development process involving Robot Operating System (ROS).

In the software demonstration, Jon and Sebastian first use a pretrained neural network in MATLAB to create a deep learning classification algorithm. Then, they use GPU Coder™ to generate a standalone library from this algorithm and deploy it to an NVIDIA Jetson™ platform. Finally, they integrate the generated library into a ROS node developed in C++ to connect with other software nodes running on the network.

Download the example files used in this video from MATLAB Central File Exchange.

Learn more with the following resources:

• NVIDIA GPU Support from GPU Coder
• Deep Learning Support from GPU Coder
• Getting started with MATLAB, Simulink, and ROS