How to Stereo Calibrate more than 2 cameras?

[aviralchharia]

Hi, I was trying 3D human pose estimation using $3$ cameras: i.e., first finding 2D keypoints from each of the $3$ camera frames and triangulating to get the 3D pose. In such a case, how to get the camera extrinsic parameters?

Specifically, suppose for $3$ cameras, should we use $[C1, C2]$, $[C2, C3]$ pairs or should we use $[C1, C2]$, $[C1, C3]$ to find the extrinsic matrix of the cameras?

How to use the GUI for the same?

[mprib]

Hi @aviralchharia. Just saw this sitting here in the issues. Sorry for the delay in noticing it. You bring up a big question. For more than two cameras you can no longer rely on something like the stereocalibration opencv functions, instead you must use a process called "bundle adjustment". This is an optimization that I'm currently implementing with the help of SciPY. It's a can of worms to get into, but hopefully just having the word "bundle adjustment" can help get you started down the google/ChatGPT rabbit hole.

In a couple weeks I'll start rolling out a rough version of this project and when I have the README updated and an intro tutorial put together, I will send a notice your way.

[mprib]

Just wanted to note that I migrated this over to discussions

[mprib]

Hi @aviralchharia! I just wanted to give you a heads up that the project is hitting the stage where I've started slowing rolling it out to folks for testing. I'm certain that there are problems in it, but it might be able to work for some of your calibration needs.

You can check out the Quick Start guide on the README and this brief walk through of a calibration and data collection here

[aviralchharia]

Thanks for Sharing!