why yolov3-tiny costs about 3 sec on the imx8qxpmek

[1127415481]

I currently test the yolov3-tiny.tflite model on the imx8qxpmek using the command "pyeiq --run object_detection_yolov3" . Surprisingly, I find it costs 3.56 sec . So I want to know whether the command "pyeiq --run object_detection_yolov3"  is running on GPU Mode or CPU Mode .dhavalMandar Can you help me?

The env on My target is as follows:

BSP version: L5.4.24-2.1.0

the command for building image: bitbake imx-image-full

eiq version: 2.1.0

the result with  the command "pyeiq --run object_detection_yolov3" :   eIQObjectDetectionYOLOV3.png

[Bio_TICFSL]

Hello du,

I guess you are using pyeIQ on CPU without using the GPU, btw PyeIQ was tested and validated only on i.MX 8MP and i.MX 8QM, we never ran tests on i.MX 8QXP, but it should work.

Regards

[1127415481]

Thank you for your reply.

I really think so. But I don't found any command or alternative parameter to indicate or select "pyeiq "  running on GPU or CPU.  Can you tell me how to  make  pyeiq run on GPU ? 

Regards