depth camera oak d lite

MohamedMagdyGaberDarwish

hello,
I am working on an object localization project using Oak d Lite camera and a custom pre-trained model on Roboflow

the camera detects the object frame but it does not show the depth it always shows a depth equal to zero
Sometimes it shows numbers like 2022, 3125, and 2865 these constant numbers do not represent the correct depth
so what I need to know is how to get the right value of the depth and why if it showed a depth value it shows a constant number
thanks

jakaskerl

Hi @MohamedMagdyGaberDarwish
I'd suggest using the SDK for this since the roboflow doesn't offer any configuration for depth/filtering.

Thanks,
Jaka

MohamedMagdyGaberDarwish

Hi jakaskerl
do u know then how can i get the three coordinates from the SDK
after i detect the target i want to print the x and y coordinates and then calculate the depth

Thanks,
Mohamed

jakaskerl

MohamedMagdyGaberDarwish
Add a callback function:

with OakCamera() as oak:
    color = oak.create_camera('color', resolution="1080p")
    model_config = {
        'source': 'roboflow', # Specify that we are downloading the model from Roboflow
        'model':'american-sign-language-letters/6',
        'key':'coFlxareOlro' # Fake API key, replace with your own!
    }

    def cb(packet: DetectionPacket):
        detections = packet.detections
        for detection in detections:
            print(detection.top_left, detection.bottom_right)


    nn = oak.create_nn(model_config, color)
    oak.visualize(nn, fps=True)
    oak.callback(nn, cb)
    oak.start(blocking=True)

Thanks,
Jaka

MohamedMagdyGaberDarwish

jakaskerl
I added the callback function but it turned out to be another error

I am sure of the connection but I don't know why it can't run

jakaskerl

Hi @MohamedMagdyGaberDarwish
Unplug and replug the device?

MohamedMagdyGaberDarwish

jakaskerl
it worked when i forced it to work in USB2, I don't know really why it doesn't read that I am using USB3. The problem may be in my laptop
Now I have another issue, I don't know what numbers the camera printing

I want to print the middle coordinate of each object detected and then the depth so do you know how it can be done ?

thanks

jakaskerl

Hi @MohamedMagdyGaberDarwish
These are coordinates for bounding box (top_left and bottom_right). You can average the two and get the centerpoint.

If you want depth as well, you need to create the model (create_nn) by passing spatial=True.

Thanks,
Jaka