Feasibility of high-frame-rate GPU image processing on Jetson Nano with Raspberry Pi HQ Camera

Developer Tools Other Tools NVAPI
, ,
1

Hello,

We are planning to use the following stack to simultaneously capture, process, and display images in real time using the GPU:

Hardware

  • NVIDIA Jetson Nano Developer Kit

  • Raspberry Pi HQ Camera

Software

  • Python

  • OpenCV (CUDA where available)

  • PyCUDA

Goal
Stream frames from the Raspberry Pi HQ camera, run GPU-accelerated image processing on the Jetson Nano, and display the output live.

We want to evaluate whether this architecture is fundamentally sound before committing further.

Question

  1. Is there any fundamental limitation (camera interface, memory bandwidth, CPU–GPU transfers, driver support, etc.) that would make this setup unsuitable for true real-time, simultaneous capture + processing + display?