Regarding multiview camera set up

[snehashis1997]

I am working in a multiview setup with 4 cameras placed at the 4 corners of a room. Based on your documentation, I followed these steps:

1. Set up the cameras

2. Calculate each camera's intrinsics

3. Run compute_relative_poses.py

4. Run concatenate_relative_poses.py

5. Run bundle_adjustment.py

6. Run triangulate_image_points.py

Can you please tell me at which step I need synchronized images? Is it required for the landmarks.json file?

[lcit]

If you are planning to estimate the 3D position of a moving object you have to time synchronize the cameras. To calibrate the cameras it is not required as long as you annotate features of a static objects. And obviously, the features must be the same in all viewpoints.

[snehashis1997]

can I used these 4 images creating landmarks.json file? these 4 images taken different 4timestamps?

[lcit]

Possibly yes. There is some overlap between the views. You want to annotate the tables, display etc but not the checkerboard. The objective is to form correspondences between at least pairs of views and to chain all views together. You will then define the chain in setup.json

[snehashis1997]

Ok, thanks
Query 1: how you are using landmark.json file, like generally sift or or like feature descriptors are used for finding corresponding points, here we are directly giving these images and with hand annotated landmark points, then there should not be any residual error?
And during marking these points is the orders matters?