openVINO python demo performance is not same as cpp demo in Yocto OS

[WT_Jacko]

Hi sirs,
I tested object detection python demo in yocto,(https://docs.openvino.ai/latest/omz_demos_object_detection_demo_python.html)
model : YoloV3 (416x416 INT8), inference performance just only 12-13 fps.
As the same HW & same yocto OS, cpp demo would be 28-30 fps.

my test environment:
TGL yocto version: release-74_tgl_u_mr3
openvino version: 2021.4
Platform : I3-1115GE
Any comments?
Thanks!
BR
Jacko

[IntelSupport]

Hi WT_Jacko,

Thanks for reaching out.

We are working on this and will update you with the information soon. Meanwhile, can I know the source of your model? Are you using YOLOV3 from OpenVINO public pre-trained models?

Regards,

Aznie

[WT_Jacko]

Hi Aznie,
Q: Are you using YOLOV3 from OpenVINO public pre-trained models?
Ans: Yes. here is my My test YoloV3 model: https://drive.google.com/file/d/1MiB5W-khMNRCSw6sB5tM5MvvUQWyLiJe/view?usp=sharing

Thanks
BR
Jacko

[Peh_Intel]

Hi Jacko,

With the same configuration, running in CPP demo is expected running faster in Python demo. This is due to the internal organization of the callback mechanism in Python API leads to a decrease in FPS. This statement is available in the Object Detection Python Demo.

Regards,

Peh

[Peh_Intel]

Hi Jacko,

This thread will no longer be monitored since we have provided an answer. If you need any additional information from Intel, please submit a new question. 

Regards,

Peh