Transfer YOLOV3 trained model from desktop to Jetson

[christian91e9l]

Hello everyone

I trained a YOLOV3 model for classification purpose on my desktop with Keras (h5 file).
I am now attempting to transfer those data on my Jetson Nano.

The Jetson Nano is working well with the demo models (Tensorrt 6.0, Tensorflow 1.15) and I am getting 20 fps with tiny288 model.
I am now struggling when attempting to transfer my own model from Desktop to Jetson.
Conversion process through ONNX format seems not to be working.
Did anyone got same trouble than me?

Christian

[AastaLLL]

Hi,

Do you have the YOLOv3 darknet format model?
If yes, you can follow our deepstream sample to run it on Jetson directly:

/opt/nvidia/deepstream/deepstream-5.0/sources/objectDetector_Yolo/

If you are using Keras file, please follow this sample to convert it into TenosrRT engine.

/usr/src/tensorrt/samples/python/end_to_end_tensorflow_mnist/

Thanks,