Make sure engine conversion works fine

[Surya-Rayala]

Also once the issues are solved verifying if the engine export works fine as well will be very helpful

[mikel-brostrom]

Please check the fixes here:

fix/engine-export

I do not have a computer with Nvidia GPU at the moment so if you help me out here, we can merge the fix faster 😄

[Surya-Rayala]

Thanks a lot mike the export works now , but now there is a small issue in inferencing with the tensorrt backend.

The Conflict
The Exporter: When you created your model, BoxMOT's exporter named the output layer output0.
The Backend: When you try to run the model, BoxMOT's inference code looks for a layer named output.
The Result
Because output0 does not exactly match output, the code crashes with a KeyError: 'output'. It is essentially looking for a folder that doesn't exist because it was saved under a slightly different name.

How to fix it quickly
You have two choices:
Edit the Backend: Open tensorrt_backend.py and change the word "output" to "output0" on line 86.
Re-export: Run the export again, but manually ensure the output name is set to just "output" (no zero).

[mikel-brostrom]

Thanks for quick feedback @Surya-Rayala. I just pushed this fix to the branch. Please try again and let me know

[Surya-Rayala]

Done I think everything work now !, Thanks mike

[mikel-brostrom]

Thanks for your help @Surya-Rayala !