Object Tracking in a script

Thor

the output of the ObjectTracking node are Tracklets. I could not find much documentation about it.
I still have not found a way to set up a pipeline that will actually use ObjectTracker to increase the frame rate to the nominal (assuming eg the camera runs at 30fps while the NN slows it down to 10)
Any working examples?

jakaskerl

Hi @Thor
standard python:

# Assuming `frame_width` and `frame_height` are the dimensions of your frame
frame_width = 1920  # Replace with actual width
frame_height = 1080  # Replace with actual height

trackletsData = track.tracklets
for t in trackletsData:
    # Denormalize the ROI manually
    roi_normalized = t.roi
    x1 = int(roi_normalized.topLeft().x * frame_width)
    y1 = int(roi_normalized.topLeft().y * frame_height)
    x2 = int(roi_normalized.bottomRight().x * frame_width)
    y2 = int(roi_normalized.bottomRight().y * frame_height)

    # Now x1, y1, x2, y2 are the pixel coordinates of the ROI
    # You can use these coordinates as needed

Thor

Thank you @jakaskerl
I figured that out, I'm glad that you confirmed it.
The issue I have, I can't get a better frame rate, which was the goal of using ObjectTracker
with detectionNetwork.input.setBlocking(False), I get 10fps from the NN and 15fps from the detectionNetwork passthrough
So I added an ObjectTracker:
objectTracker = pipeline.create(dai.node.ObjectTracker)
objectTracker.setDetectionLabelsToTrack([0])
objectTracker.setTrackerType(dai.TrackerType.SHORT_TERM_IMAGELESS)
objectTracker.setTrackerIdAssignmentPolicy(dai.TrackerIdAssignmentPolicy.UNIQUE_ID)
objectTracker.inputTrackerFrame.setBlocking(False)
objectTracker.inputDetectionFrame.setBlocking(False)
objectTracker.inputDetections.setBlocking(False)

but I get 10 fps from the ObjectTracker outputs, which is the still the same NN fps
Any help?