I use just 30 feature matching point sets here. I do not use interp2 because I think it is too slow. Instead I just use round to find the nearest neighbor for inverse warping, so the quality of the result might not be very good. Also I use a sine function for the cross-dissolving factor so that it can get rid of the shadows out of the shape triangulation faster while the warping factor is still linearly interpolated. I think the result video looks better, but the middle of the video isn't really the mean face of the two original faces.
Frames in the original size and the uncompressed morphing AVI video can be found here
Mean Face:
Nothing much to do here, the hardest work here is to create the point set data for every one in 15-463. After that just repeat warping images and then finally average them together to get the mean face.
Mean face of 15-463:
My face warped into mean shape:
Mean face warped into my shape:
Bells & Whistles: Caricatures
There are not enough data to make the caricatures more accurate, so I cannot tell significant differences here, or maybe my face is too close to the mean face. But there are still some noticeable changes in the pictures.