Question about 3D pose estimation

Dear all,
I am trying to use deeplabcut for 3d pose estimation on some random animal videos that are downloading from the internet.
When I reached the camera calibration step, I had some questions.

  1. Since the video is downloaded from the internet, I could not take pictures of the checkboard s using the camera that took the video. Is there any way that I could do the camera calibration simply using the original video? I checked the cheetah example and I didn’t fully understand the implementation details of the camera calibration. It seems we still need the pictures of the checkboard.

  2. If the video itself is animated, does this mean that we can’t use deeplabcut for 3D pose estimation on this video?

  3. Could deeplabcut do 3D pose estimation by using one view?

Thank you,
Frederica