I'm trying to use Optical Flow in my pipeline to estimate conveyor belt velocity.
If I set motion estimator type to LUCAS_KANADE_OPTICAL_FLOW like this, it works.
featureTrackerConfig.motionEstimator.type = dai.FeatureTrackerConfig.MotionEstimator.Type.LUCAS_KANADE_OPTICAL_FLOW
But I see a crash if I set it to hw acceleration and then try to get features
featureTrackerConfig.motionEstimator.type = dai.FeatureTrackerConfig.MotionEstimator.Type.HW_MOTION_ESTIMATION
[system] [critical] Fatal error. Please report to developers. Log: 'opticalFlowPipe' '356'
Can you please help with this?
To add more info, I am using my OAK D lite for object detection using MobileNetSSD already. I tried to add optical flow feature to the same camera so that I can use both sequentially.
For feature tracker, I currently set numShaves = 2 and numMemorySlices = 2 or 3
Since I have object detections also enabled, I see this warning in the beginning:
[DetectionNetwork(5)] [warning] Network compiled for 6 shaves, maximum available 10, compiling for 5 shaves likely will yield in better performance
But I thought this is trivial and ignored it