Acquisition of information about objects detected with custom detectors

At the moment ,I have customized a detector for yolov8 according to the tutorial and it is recognized. I would like to know if it is possible to quickly obtain information about the detected objectsobject, including dimensions and velocity.

Hi @NineSeven
when you retrieve the detected object as an ObjectData type you can extract all the available information:
https://www.stereolabs.com/docs/api/classsl_1_1ObjectData.html

When I obtained the dimensions of the object to be detected, I found that the width and length of the data were the same regardless of the viewing angle, which is obviously very different from the truth, do you know what the reason is?

@NineSeven Can you explain what you mean?
Pictures showing the problem can help understand it better

The width and height of the object I actually measured were 5.9, 5.5, and 14.6 (cm) respectively, and now the output is as follows(id is 10):
屏幕截图 2023-11-10 162938

Do you have pictures of the objects showing bounding boxes and measures to understand what’s happening?
Numbers without a visual context are not enough to analyze the problem.


It seems that you are too close to the object and you are using a depth mode that does not provide good point cloud information.

What settings are you using?

Please also note that the bounding box of the object is not oriented according to the object, but it’s always perpendicular to the image sides, so different orientations of tall objects can lead to different measurements.

Currently, I want to measure the size of sea cucumbers with a zed camera, is there any good way to do it?

Hi @NineSeven
there are many possible ways to achieve your goal.
You can process the point cloud information, you can process color information and then project the result to 3D, you can use AI, …
The appropriate method of selection depends on your individual skill set.