| abstract |
An object or body-mounted camera apparatus for recording surgery is provided that is adapted for tracking a relevant visual field of an on-going operation. To help maintain visibility and/or focus of the visual field, specific machine learning approaches are proposed in combination with control commands to shift a physical positioning or a perspective of the camera apparatus. Additional variations are directed to tracking obstructions based on the visual field of the camera, which can be utilized for determining a primary recording for use when there are multiple cameras being used in concert. |