Hello, I imported the trained YOLOv8 model into the OAK-D-Lite camera by referring to roboflow's tutorial on how to import the model onto the Luxonis Oak device. The reference link is as follows: https://docs.roboflow.com/inference/luxonis-oak Now I am having some problems and I would like to ask you what I should do. In the colour image I can see the bounding box of the recognised object. And in the command line output of the pycharm software, I can see the following message: PREDICTIONS [{'x': 256.0, 'y': 294.5, 'width': 188, 'height': 293, 'depth': 50.91346153846154, 'confidence': 0.74462890625, 'class': 'wipes'}] [URL=https://i.imgur.com/C44OTAL.png][IMG]https://i.imgur.com/C44OTAL.png[/IMG][/URL] But in the displayed depth image I can't see the bounding box of the detected object. Do you have any suggestions if I want to show the bounding box of the detected object? [URL=https://i.imgur.com/DPlIGbT.png][IMG]https://i.imgur.com/DPlIGbT.png[/IMG][/URL] I also have another problem that the position of the target object does not change, but the depth data detected changes every time, what is causing this? [URL=https://i.imgur.com/arOnlVs.png][IMG]https://i.imgur.com/arOnlVsh.png[/IMG][/URL] I am looking forward to your reply. Thank you very much!

Hi @"Jalen_Xu"#p6892 , Instead of the roboflow oak, which is very limited, I'd suggest using our depthai SDK integration with roboflow. [Example here](https://github.com/luxonis/depthai/blob/main/depthai_sdk/examples/NNComponent/roboflow_integration.py), and their blog post: https://blog.roboflow.com/deploy-roboflow-model-luxonis-depth-sdk/ With that, you can easily get 3D coordiantes of the object, and also draw bounding boxes on the depth frames. Thoguths? Thanks, Erik

Training YOLOv8 model with the Roboflow

Jalen_Xu

Hello,

I imported the trained YOLOv8 model into the OAK-D-Lite camera by referring to roboflow's tutorial on how to import the model onto the Luxonis Oak device. The reference link is as follows:
https://docs.roboflow.com/inference/luxonis-oak

Now I am having some problems and I would like to ask you what I should do.
In the colour image I can see the bounding box of the recognised object. And in the command line output of the pycharm software, I can see the following message:
PREDICTIONS [{'x': 256.0, 'y': 294.5, 'width': 188, 'height': 293, 'depth': 50.91346153846154, 'confidence': 0.74462890625, 'class': 'wipes'}]

But in the displayed depth image I can't see the bounding box of the detected object. Do you have any suggestions if I want to show the bounding box of the detected object?

I also have another problem that the position of the target object does not change, but the depth data detected changes every time, what is causing this?

I am looking forward to your reply. Thank you very much!

erik

Hi Jalen_Xu ,
Instead of the roboflow oak, which is very limited, I'd suggest using our depthai SDK integration with roboflow. Example here, and their blog post: https://blog.roboflow.com/deploy-roboflow-model-luxonis-depth-sdk/

With that, you can easily get 3D coordiantes of the object, and also draw bounding boxes on the depth frames.
Thoguths?
Thanks, Erik