unable to compute spatial coords while using deepsort tracking example

Ahmedrashad186

Hello, i’m using example gen2-deepsort-tracking to track people and perform re-id as specified by deepsort. I want also to compute the depth of each person tracked its Z coordinates. I want to know how to integrate the depth to this example as it uses Oakcamera class not the normal pipeline

lovro

Hi,

To get the Z coordinate (depth) of each tracked person, you need to enable stereo depth and read the depth at the center of each detection’s bounding box. (Or take an average of a few pixels to make it better)

Since this example uses the SDK, the best approach is to use a callback (e.g. on tracked detections), fetch the latest depth frame, and sample the depth at the center of the bounding box.

The DepthAI SDK provides a way to do this via its trigger/callback system — for reference, see this: Custom Trigger.

Hope this helps!

Ahmedrashad186

hello, thank you for you response

i'm already enabling spatial=true on create-nn so i think yolo detections have spatial coords calculated too so can i access them from the detection info directly ?

lovro

Hi, yes you're right — if you pass spatial=True when creating the YOLO node, DepthAI-sdk will provide spatial data automatically. You can access the Z (depth) value like this:
det = packet.detections[track.detection_id]

z = det.img_detection.spatialCoordinates.z # Depth in millimeters

This gives you the distance from the camera to the detected object.