This is produced by inputting images images of our faces with a common point-set, finding a weighted mean point-set, triangulating on that set, then finding point-sets at 61 different intervals between our faces. Each face is then morphed to that point-set and blended by the percentage between each face it is. The morph is executed by placing each coordinate on the face in a triangle, then converting it to barycentric coordinates on the starting triangle then back-converting it on the triangle of the mean point-set.
This face consists of all faces in the class averaged together by morphing each of them to their average point-set then averaging all them together.