Running ONNX models on Arduino UNO Q GPU/NPU for faster inference?

Official Hardware UNO Family UNO Q
1

Hi everyone,

I'm working on a remote wildlife monitoring project and looking into using the Arduino UNO Q to run a compact/pruned version of the MegaDetector model for edge detection.

Because CPU inference might be a bottleneck for real-time video/photo processing, I'm hoping to leverage GPU acceleration.

Has anyone successfully run ONNX models offloaded to the Qualcomm Adreno GPU?

If anyone has a minimal working example or repo showing ONNX GPU execution on the UNO Q, I'd really appreciate a link or any pointers!

Thanks!

2

Hi @beaglec. The QRB2210's quad-core Cortex-A53 CPU should be best suited for this application.