Re-training SSD-Mobilenet question

Can use a custom model created in this guide here in python? If yes can write an example please.

Hi @EpicMario71, if you custom detection model was trained with the PyTorch code from the tutorial (train_ssd.py), then yes you can run the model through Python with detectnet.py. It accepts the same command-line arguments that are shown in the guide you linked to.

is there any way to load the model into the python code without adding arguments? Write me an example please.

Yes, like this:

net = jetson.inference.detectNet(argv=['--model=path/to/your/model.onnx', '--labels=path/to/your/labels.txt', '--input-blob=input_0', '--output-cvg=scores', '--output-bbox=boxes'])
1 Like