Health Check
We are struggling with clarifying an interpretation on the health check.
We are performing the calibration as described in the RealSense™ Self-Calibration for D400 Series Depth Cameras : Addendum A (March 2022): Tare Calibration with Ground-Truth Target
According to the health check it is good < 0.25.
This however does not seem to apply to the tare calibration since the health check value is given in percentage. What would be in that case an acceptable value for the health of the tare calibration?
-
Hi Maren Pedroza The 0.25 health check score that you experienced applies to On-Chip calibration, whose purpose is to improve depth image quality. Tare calibration is focused on improving depth measurement accuracy and does not have the same health-score guidance.
The accuracy of the Tare process will be dependent on providing an accurate ground truth value, which can be done automatically in the RealSense Viewer's Tare calibration tool by clicking the 'Get' button once the camera has been positioned in front of the target image print-out at the desired distance from the target.
-
Hi MartyX Grover, thank you for the clarification. Is there a threshold we can refer too? How much percentage is still good?
-
I am not familiar with this Tare health check, though clearly Tare does have its own score and so I have edited my earlier comment to reflect this. My apologies for the confusion.
The calibration documentation does not provide guidance about interpreting the Tare health score. My interpretation of it though is that so long as you have selected the High Accuracy Visual Preset in the Viewer's options side-panel before activating the Tare calibration and have used the Get button to auto-calculate the ground truth value then there is not much that can go wrong with the Tare process.
-
If the 'closer to zero the better' principle of the On-Chip calibration also applies to Tare (the documentation does not state this) then if the old calibration was 0.0570 and the new calibration was less than 0.0570 - such as 0.0350) then you could assume that the smaller value was a better calibration.
Please sign in to leave a comment.
Comments
7 comments