search

marcoslucianops / DeepStream-Yolo

NVIDIA DeepStream SDK 7.0 / 6.4 / 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 / 5.1 implementation for YOLO models
MIT License
1.39k stars 344 forks source link

Impossible to get detection in yolov8 #411

Open hdnh2006 opened 11 months ago

hdnh2006 commented 11 months ago

Hello,

Thanks for the fantastic job you have done with this repo.

In the past, I could succesfully export and run an app using deepstream in my Jetson Xavier NX (https://github.com/marcoslucianops/DeepStream-Yolo/issues/390#issuecomment-1632510589).

However, no matter what I do, what version of onnx or opset I use, I can't see the bounding boxes in the image predicted.

These are the settings I have tried (always torch==1.12):

I am always using the same command: python3 export_yoloV8.py -w yolov8s.pt --dynamic --simplify --opset **

I am leaving all the weights I have used in this link (available one week).

Can somebody help me to understand what I am doing wrong? maybe something is being wrongly cached? I don't understand why I could do it in the past, and what I am doing different now. I have even checked the .bash_history and I am repeting the same process and nothing...

Please help.

marcoslucianops commented 11 months ago

What's your DeepStream version?

hdnh2006 commented 11 months ago

Sorry @marcoslucianops for the late reply.

I am using deepstream 6.2 on my Jetson Xavier NX device.

marcoslucianops commented 11 months ago

I don't have a Jetson board to test it now. But you can export the ONNX model in your computer and use on the Jetson. Can you try?