Hello, I am trying to use an OAK FFC 4p with two IMX378 cameras as a stereo camera. After mounting them on a custom 3D printed piece, I managed to calibrate the stereo pair following this guide : https://docs.luxonis.com/projects/hardware/en/latest/pages/guides/calibration.html I got the following reprojection and epipolar errors : [upl-image-preview url=https://discuss.luxonis.com/assets/files/2023-05-16/1684244041-559023-capture-decran-de-2023-05-16-15-31-54.png] (I am using the latest release btw : https://github.com/luxonis/depthai/releases/tag/v1.10.0-sdk) I had to bypass a status check to force the flashing of the eeprom, as I got an average epipolar error of 0.9 and you say in the documentation : "An average epipolar error of

Hi @"apirrone"#p8323 , > "An average epipolar error of

OAK FFC 4p with two IMX378 depth disabled when runnning depthai_demo.py

apirrone

Hello,

I am trying to use an OAK FFC 4p with two IMX378 cameras as a stereo camera. After mounting them on a custom 3D printed piece, I managed to calibrate the stereo pair following this guide : https://docs.luxonis.com/projects/hardware/en/latest/pages/guides/calibration.html

I got the following reprojection and epipolar errors :

(I am using the latest release btw : https://github.com/luxonis/depthai/releases/tag/v1.10.0-sdk)

I had to bypass a status check to force the flashing of the eeprom, as I got an average epipolar error of 0.9 and you say in the documentation : "An average epipolar error of <1.5 is considered to be good, but not required.", but the threshold must be lower than 0.9.

Anyways, the flashing seems be OK :

However, when I run the depthai_demo.py script, the depth (along with other things) is disabled :

Any idea why ?

Thanks !

jakaskerl

Hi apirrone
Afaik, the depthai demo doesn't yet support RGB stereo for FFC devices. It doesn't detect the expected mono cameras and therefore disables depth. You will have to manually run both cameras through stereo node to get depth/disparity preview.

Hope this helps,
Jaka

apirrone

Ah ! I did not know, thanks !

erik

Hi apirrone ,

"An average epipolar error of <1.5 is considered to be good, but not required."

Where in docs did you find this? Just so we can update it🙂
THanks, Erik

apirrone

Hi @erik It was in the parse_args() function of calibrate.py in the depthai repository.

While I am at it, I also noticed that the default values of some parameters do not match their written description 🙂

erik

Hi apirrone ,
I agree calibration.py is far from perfect - as such calibration is rarely used (we use TV calibration in factory, quite different code). Hopefully we will be able to update it soon when it will be integrated into the depthai viewer directly (GUI altnerative to depthai_demo.py). More info on it in the upcoming week(s).