To find the mid-way morph, the average of the two images' correspondence points were used to find a Delaunay triangulization. Next, both images were warped to the average shape and the two warped images were cross-dissolved together (using weights of 0.5 and 0.5) to find the mid-way morph.
To generate a morph sequence, the values of warp_frac and dissolve_frac were incrementally changed from 0 to 1.
Using the previously described morph function, an average face of a subset of the Danes dataset in pose 2 could be determined. This was accomplished by calculating the average shape of all the faces and warping each face to the average shape.
Original Face
Warped Face
Probably for the best that I am not Danish
Looks mildly racist but I guess that is my face shape ¯\_(ツ)_/¯
Caricature images could be created by extrapolating from the mean average. This was done by first taking the difference between my face's and the average face's correspondence points. The difference was then scaled by a value, alpha, and used to exaggerate my features
Jodi and the world's most average white male
Jodi's face warped to the geometry of an average male
Jodi and the average male morphed
I used to want a brother, but maybe not anymore...