imx8: running pyeiq facial expression detection on 5.4.24 bsp with video source thowing error

[vsuneja63]

I could run facial expression detection on imx8mq evk(5.4.24 bsp) with input image. Same i tried to run with video source but its thowing error. Here are the logs:

INFO: Created TensorFlow Lite delegate for NNAPI.
Applied NNAPI delegate.
Using /dev/video2 as video device
Resolution not supported. Using 640x480 instead.
[ WARN:0] global /usr/src/debug/opencv/4.2.0.imx-r0/git/modules/videoio/src/cap_gstreamer.cpp (713) open OpenCV | GStreamer warning: Error opening bin: no element "imxvideoconvert_g2d"
[ WARN:0] global /usr/src/debug/opencv/4.2.0.imx-r0/git/modules/videoio/src/cap_gstreamer.cpp (480) isPipelinePlaying OpenCV | GStreamer warning: GStreamer: pipeline have not been created
Your video device could not be initialized. Exiting...

As per my understanding imx8mq evk doesn't support g2d. So what can be done here to make it working?

Would like to here from you asap.

[Alifer_Moraes]

Hello, vsuneja63@gmail.com‌,

PyeIQ's demos use v4l2 pipelines as default to stream the video captured from your camera, and those pipelines use imxvideoconvert_g2d for video conversion. Did you specified any --video_fwk? Could you please try to use --video_fwk=opencv and see if it works?

Regards,

Alifer

[vsuneja63]

Hi

Thanks!

I tested with command "pyeiq --run face_and_eyes_detection --video_src=/path_to_the_video". I didn't specify "--video_fwk=OpenCV" in this, will test with this & update on the same. Does g2d is used in codebase? As per my understanding g2d is not available in imx8mq evk.

[Alifer_Moraes]

Hello,

Yes, we are using g2d in the code of our v4l2 pipelines, since we developed PyeIQ for i.MX 8MP and i.MX 8 QM. To run the demos on i.MX 8MQ try to use --video_fwk=opencv.

Regards,

Alifer

[vsuneja63]

Hi 

Thanks!

It worked by adding this with command. Just for my info what exactly this flag does here.

[Alifer_Moraes]

Hello,

The --video_fwk flag is used to choose which video framework is going to be used to stream video, the available options are v4l2 (default), opencv and gstreamer (experimental). Both v4l2 and gstreamer run with hardcoded pipelines that use g2d and opencv is a simple default pipeline with minimal configuration to run in most environments.

Regards,

Alifer