Mediapipe/TensorFlow Lite GPU Delegate on Jetson Orin Nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
, ,
1

Hey y’all! I’m trying to use OpenCV hand landmark tracking on my Jetson Orin Nano using mediapipe, which I built using the following tutorial: How to Download & Build MediaPipe on NVIDIA Jetson Xavier NX?

My test python script runs and the landmark tracking does indeed work, however I am running at <5 FPS. Reading the terminal states that TensorFlow Lite is initializing with the CPU on the Orin Nano as the delegate and not the GPU. OpenCV does recognize the CUDA device, but I’ve been slamming my head against the wall trying to figure out how to configure TensorFlow Lite to use the GPU as the delegate.

Any suggestions on how to do this? Or maybe any other options for hand position tracking on the Jetson Orin Nano?

Thanks guys!

3

Hi,

You will need to install the package that has been built for the CUDA support.

For OpenCV, below is an automatically building script for your reference:

For TensorFlow (not the lite version), please find the document below:

Thanks.

5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.