YOLOv8 / ONNX object detection, multiple shapes on top of eachother

[mfwre]

Sorry if this is a dumb question / issue. I'm really new at machine learning / object detection.

I exported a YOLOv8 nano model to ONNX and ran object detection with your runtime and with the ort crate.

Whilst with the ort runtime I get one shape each, with yours I get several same object detected stacked on top of each other - same middle point, almost same width and height.

Is this related to the runtime? Do I have to clean up the output myself?

I hope I expressed myself correctly.