Hello,
I am trying to create a FMO counting program, using the OAK-D Mono Camera.
The trouble is even with the global shutter on the Mono Camera I still get extremely blurry images like so:
Is this normal behavior or is there any way to mitigate or stop this directly in the camera configuration?
Kind Regards,
Luiz Sutil
luiz_sutil Reducing the exposure time is likely what is needed here. As global shutter prevents tearing/jello in the images - as all the pixels are captured at the same time.
But if the exposure time is occurring over a period in which the object of interest has significant motion, the capture will have this motion in it.
So generally the solution is to reduce the exposure time while increasing lighting as necessary to keep adequate brightness for the application.
And example of controlling exposure and other parameters on the mono (grayscale) cameras is below:
https://docs.luxonis.com/projects/api/en/latest/samples/MonoCamera/mono_camera_control/
Thoughts?
Thanks,
Brandon
Got it working! I still have some setup to do but it had a very clear effect!
Another question, my detections all seem to lag a frame or two, like so:
I found a few examples that sync the NN with the color camera, but couldn't get it working with the mono. Could you point me to some direction on what the best syncing method for my use case would be?
Thank you very much for the assistance!
Hello luiz_sutil , HostSync class here. In the readme of that demo, you can see it's perfectly synced, so blurring happens on the face, not behind it. Let me know if you have any trouble adding this helper class to your project.
Thanks, Erik
Setting both the NN and frame output queue solved my problems after a bit of tinkering! One last question, the NN detection from the MobileNetDetectionNetwork is occurring on the device (OAK-D) correct?
luiz_sutil correct, both MobileNet inferencing and decoding are done on the device (OAK-D) itself.
