StereoPi Forum

I'm using the stereopi set up, and created my own script using opencv for calibration. The chessboard corners are detected very well in the left camera images each time, but the right camera images struggle to correctly identify chessboard corners. Below is the link to my github with the code. I have been using the images in the "pairs_land_calib" folder as well as the script, cv_calibrate_stereo.py

https://github.com/atrunyan/stereopi_URIL.git

I did not test your code, but by reading it rapidly, I noticed an error here:
https://github.com/atrunyan/stereopi_UR ... reo.py#L56
Not read afterwards.

Thank you! I actually caught this same error shortly after posting my issue. It is now fixed!

I am still not getting a well-defined disparity map though. My maps are all very noisy, there is no clear shapes even for a basic picture of a box against a blank wall.

Hi atrunyan,

I've spotted your "cutted" pairs looks swapped, i.e. left image is actually a right image.
Examples:
https://github.com/atrunyan/stereopi_UR ... eft_01.png
https://github.com/atrunyan/stereopi_UR ... ght_01.png

Could you please check this part of your pair-cutting code?

Also I can't realize either you are using a wide-angle camera, or the chess-board is not a plane? ( I guess it's a widfe-angle with specific optics...)