Help with converting FasterRCNN detection models from Pytorch

[aleda86bbl]

Hi there,

I am trying to convert an ONNX model that I had previously converted from a PyTorch model; the FasterRCNN (fasterrcnn_resnet50_fpn) model that is found in torchvision.

Unfortunately the conversion is unsuccessful and I get an exception:

raise NotImplementedError('Dynamic value of min/max is not implemented') from exc NotImplementedError: Dynamic value of min/max is not implemented

I was wondering if anyone has had any luck in converting this model back to PyTorch?
Since it was converted some time ago, I tried converting the opset version from 14 to 17 but that didn't make a difference.

Any tips or suggestions would be greatly appreciated,
Alexander