Hey guys I just bought the OAK-D PRO. On the factory setting I ran this SpatialOnHostDemo, where I was standing in a straight line from the camera at a distance of 60 inches (1.5 m) where I was getting a error range of 5 inches. Im pretty new to using this camera and the depthai as well. Can someone please help me if I need to change the camera configurations or my approach on getting the depth measure to reduce the error to less than 1 inch. Any suggestions are highly appreciated!
OAK-D PRO Depth accuracy.
Prinal
Can you share the MXID of the device so I can check factory calibration?
Also try using useSpec=True here - luxonis/depthai-experimentsblob/master/gen2-calc-spatials-on-host/calc.py#L47
Thanks,
Jaka
Thank @jakaskerl. I tried changing the useSpec=True, but that still generated ~5 inches of error like the previous runs. My devices MXID=14442C1011D4D2D600. Thanks for helping on this.
Hi Jaka,
Im using this SpatialOnHostDemo. All Im doing is plugin in the OAK-D Pro, run the demo and see the measurement on the screen from the object (me standing in the front of the camera). Then when I measure the distance from the camera to the me standing at the same position using a measuring tape, Im getting a error range of 5 inches. Which I think is significant when Im standing only 60 inches far from the camera.
Prinal
What ROI is taken? Is it overlaid correctly so it doesn't calculate the background depth as well?
https://docs.luxonis.com/software/depthai-components/nodes/mobilenet_spatial_detection_network/#MobileNetSpatialDetectionNetwork-Common%20mistakes
Thanks,
Jaka
Hi Jaka,
Yeah while running the SpatialOnHostDemo, we kept the ROI entire intact to the object. But still the error range of ~5 inches is showing. Are there any other configuration changes apart from the one you suggested we need to try?