Deploying Deep Neural Networks to Embedded GPUs and CPUs

Designing and deploying deep learning and computer vision applications to embedded GPU and CPU platforms like NVIDIA^® Jetson, AGX Xavier™, and DRIVE AGX is challenging because of resource constraints inherent in embedded devices. A MATLAB^® based workflow facilitates the design of these applications, and automatically generated C/C++ or CUDA^® code can be deployed to achieve up to 2X faster inference than other deep learning frameworks.

This talk walks you through the workflow. Starting with algorithm design, the algorithm may employ deep learning networks augmented with traditional computer vision techniques and can be tested and verified within MATLAB. Bring live sensor data from peripherals devices on your Jetson/DRIVE platforms to MATLAB running on your host machine for visualization and analysis. Deep learning networks are trained using GPUs and CPUs on the desktop, cluster, or cloud. Finally, GPU Coder™ and MATLAB Coder™ generate portable and optimized CUDA and/or C/C++ code from the MATLAB algorithm, which is then cross-compiled and deployed to Jetson or DRIVE, ARM^®, and Intel^® based platforms.