The Multi-Cam system requires a camera calibration. Using this, it can automatically determine the bone lengths of actors during the action takes, without requiring any actor calibration. This means that once a camera calibration has been done, anyone else can do the action takes without needing to do any preparation!
To Record a Calibration:
Hit record.
Stand in the centre of the capture area.
Clap 3 times above your head clearly, ensuring it is sufficiently louder than any background noise. (Note - this isn’t necessary if your cameras are synchronized)
Stand still in a Y pose for 2 seconds, then, still holding the Y-pose, walk from the middle of the capture area right to the edge of the volume in front of each camera, facing the respective camera at all times. Move in a natural motion, pause at each camera and return to the centre while still facing the camera.
Ensure that you walk towards each camera, until you are as close as possible whilst still fully framed. This will ensure the area that you calibrate will represent the area that you will do your performance within.
Repeat this motion walking to each camera, then backwards to the centre, before moving on to the next.
If using less than 4 cameras, walk around the perimeter of the volume to ensure the system has an accurate understanding of the perimeter of the capture volume. (The perimeter you define, will be used to ensure only the desired actors are tracked during action takes)
Now finish the recording.
At this stage, we recommend measuring the height of your actor and noting it down, you'll need it later when you process the calibration! Their height should not include the footwear they are wearing.
We always recommend doing an extra calibration at the end of your session, just in case a camera was moved during the session.
Check Out This Example!
Note - this example calibration was done using a synchronized camera system so the actor didn't need to clap. If you are using cameras that aren't synchronized, you'll need to do 3 claps at the beginning of the video.
In this example, you would enter the following sequence times:
No clap times (these cameras are sync'd)
Y-pose sequence start time: 00:08
Y-pose sequence end time: 00:37
Additional notes:
Make sure only the actor (i.e no other people) are in the capture volume, or holding a Y-pose during the calibration otherwise the system may track the wrong actor.
If the cameras are moved (even slightly) at any point during the shoot, a new calibration will be required for optimal results.
What is the purpose of the calibration?
The calibration process provides the system with a few crucial pieces of information:
The positions and orientations of your cameras:
The system needs to determine this information about the cameras so that it can correctly interpret the location of the actors seen in the footage during the action takes.
The scale of the 3D worldspace:
By knowing the height of the calibrating actor, the system can determine their position by triangulating their keypoints in the footage from each camera. Based on this, the system will be able to automatically determine the size and position of the actors seen in the action takes.
The origin on the 3D worldspace:
When you process your calibration, you enter the start time of the Y-pose sequence. The system use the location and orientation of the actor at this time, to lock in the location and orientation of the origin. The animations for all action takes using this calibration will then be positioned with respect to this origin
The perimeter of the capture volume (check out the below visualisation):
What - The perimeter of the capture volume is used to determine who is tracked during your action takes. When you process an action take, you will enter the start and end time and the number of actors to be tracked. In the first frame of this window, the system will search for the given number of actors within the perimeter, to ensure that only the desired actors are tracked. If the desired actors are stood outside this perimeter at this time, they will not be tracked. If any actors enter this area after the start time, they will not be tracked.
How - The perimeter is created automatically, determined by the furthest locations the actor moves from the origin. As the actor walks towards each camera, these locations will be logged and will be used to define the perimeter of a convex hull, otherwise known as the capture volume. If less than 4 cameras are used, we recommend walking around the perimeter of the volume after the usual sequence (still in a Y-pose), so that the perimeter has enough keypoints to ensure it doesn't cut off any corners of your capture volume.
