Using three-dimensional scans of a physical subject to determine positions and/or orientations of skeletal joints in the rigging for a virtual character

A method comprising:

• receiving a plurality of polygon meshes of a digital character, the polygon meshes being representative of a physical subject in a plurality of training poses, each training pose comprising a unique articulation of one or more body parts of the physical subject;
• segmenting each of the plurality of polygon meshes corresponding to the plurality of training poses into a plurality of articulation segments;
• determining an indicator of position or orientation for each of the articulation segments for each of the polygon meshes corresponding to the plurality of training poses;
• determining, based on the indicator of position or orientation of the each of the articulation segments, the position or orientation of one or more joints for the plurality of training poses;
• and
• creating, based on the position or orientation of the one or more joints for the plurality of training poses, a pose space for a pose space deformer to calculate mesh deformations for the digital character by interpolating between the plurality of training poses,
• wherein determining an indicator of position or orientation for each of the articulation segments of the polygon mesh comprises determining characteristic axes for each of the articulation segments, a first characteristic axis pointing in the direction of greatest spatial variation between the set of vertices in the articulation segment, a second characteristic axis pointing in the direction of greatest spatial variation between the set of vertices in the articulation segment, subject to the constraint that the second characteristic axis is perpendicular to the first characteristic axis, and a third characteristic axis being perpendicular to both the first characteristic axis and the second characteristic axis.