Unable to load Custom Trained Yolov8m with streaming example

General
, ,
1

I have compiled my custom trained Yolov8m to HEF format following this guide https://community.hailo.ai/t/creating-custom-hef-using-dfc-model-zoo/739

Then I tried to run it with Streaming example from [Preformatted text](https://github.com/hailo-ai/Hailo-Application-Code-Examples/tree/main/runtime/python/streaming)

But the code cannot load the model properly.
On this line

image
image853×101 8.54 KB

The infer_result.keys() returns:
image

While if I load the original model (yolox_s_leaky.hef), I got this:

image
image1544×66 3.96 KB

Am I missing something during the optimization and compiling?

Here’s my alls file:

image
image793×166 5.25 KB

And my nms_config:

image
image406×605 6.46 KB

Everything is complied in Ubuntu20.04 machine, with:

  • hailo-dataflow-compiler==3.28.0
  • hailo-model-zoo==2.12

But the inference is on Windows 10 machine, with:

  • hailort==4.18