High CPU usage when running eight on-board leopard imaging cameras with Orin 64GB

Our team has been developing the Nvidia Nova Carter robot that was obtained from Segway, which has eight onboard Leopard Imaging cameras (four Hawk 2D cameras and four Owl cameras). However, we face large frame drops and high CPU usage on the Orin 64GB Dev board when employing them. More specifically, we obtain approximately ~100% CPU usage per camera (running teleop_launch.py without the lidars) and achieve only one image frame every few minutes, which is unacceptable for our use case. Here, we note that we have strictly followed Nvidia’s instructions on the set-up, and the OS is flashed on the SSD (Samsung 980 Pro).

Please share your experiences with achieving the eight cameras running simultaneously with the Orin 64GB.

More specifically, could you share with us the following details:

  1. Jetpack version
  2. The driver version for Leopard cameras
  3. Parameters of the cameras (resolution, etc)

Your assistance would be greatly appreciated.

Thank you.

Yours sincerely,
Zhi Wei

Hi @zhiwei001

Welcome to the Isaac ROS forum. I forward your issue to our engineers for more details on your issue