How to use real-time object detection with tensor RT?

[olivier.berton.leclercq]

Hi,

I’m new to AI and computer vision. I use a jetson nano. I trained myself with YOLO on darknet with a laptop and hello AI world with the jetson nano.

If I understood well, TensorRT can help models to be better on jetson nano.
Let’s say I would like to use YoloV3 from darknet, what do I excactly have to do?
I suppose I have to convert somes files?
And after, I converted those files, which command on the jetson nano let me use my program? Or do I need a framework?
During the tutorial hello AI world it seemed that a framework wasn’t necessary because of tensorRT?

Sincrely,

[dusty_nv]

Hi @olivier.berton.leclercq, the jetson-inference project doesn’t explicitly support importing YOLO models from DarkNet. However, there is a TensorRT sample that supports YOLOv3 included with JetPack 4.3 - you can find it under /usr/src/tensorrt/samples/python/yolov3_onnx