YoloX problem

liuxinze

Hi there,
I have heard Yolox algorithm is better than Yolov5. Therefore I want to deploy YoloX on my OAK-D. I try to run depthai-experiments/gen2-yolox program with my own data. I have trained my data and convert the .pt file to .onnx file, then to .xml and to .blob file when I exactly follow the steps in the Readme.md.
However, the output preview has not detected any objects although the program is running. There is no errors displayed but no detected objects is shown. So I hope someone can help me solve it.
Thank you,

liuxinze

erik

Hello liuxinze , I have asked our ML engineer to write some suggestions. I would also suggest asking this question on our Discord server, as there's the whole community that can provide suggestions/help as well🙂
Thanks, Erik

Matija

Hey, could you please provide a bit more info about the model?

Which version did you use (nano, tiny, S)?
What's the input shape?
What have you changed in the main.py?
Was the model working before exporting it to OpenVINO?

Also, for faster response I'd suggest the same as Erik - join our Discord channel as we have an active community which might be able to help you before we do 🙂