モーションキャプチャーのパイプライン

Kinoca's motion capture pipeline works roughly in 4 steps:

1. Detection of bounding box for all visible actors in each camera

2. Extraction of 2D keypoints (features) for the actor in each camera

3. Triangulation of all keypoints into 3D with respect to the global frame

4. (optional) Perform inverse kinematics (IK) to convert the 3D keypoints into bone transforms for a given 3D model

注釈

Kinoca currently supports only single person tracking, although multiple people tracking is a planned feature.