Num_det always zero

[tuananhroman]

Env

• GPU:
  GeForce GTX 860M
• OS:
  Ubuntu 18.04
• Cuda version
  11.0
• TensorRT version
  8.0.0

Hey everybody,

I'm currently working on a panorama live tracking application and intended on speeding up the inference with TensorRT. For engine generation, I'm using the exact same code as in this repo. Subsequent inference always returns zero though,
as num_det equals zero on every instance even though there is some reasonable result in the outputs array for outputs[0].

How do I need to change line 165 in order to retrieve the true number of detections found?
Or is there a fundamental error in the respective engine?

Thanks a lot!