Defining Correspondences

I wanted to morph my face into one of my favorite actors: Dev Patel.

I started by defining the following correspondence points for both faces:

And then I computed the delaunay triangulation of the midway average face shape. The triangulation resulted in the following:

Computing the "Mid-way Face"

After that I computed the midway warp of both images and averaged the colors to get the following result:

The Morph Sequence

In order to create the morph sequence, I created a total of 45 frames where each frame had ascending warp and cross dissolve fractions between 0 and 1.

The "Mean face" of a population

Next I computed the Mean face of the Danes dataset. First I computed the average face shape of all the Danes. Then I morphed each of the Danes' faces to the average face shape. Below are a few examples of Dane faces being morphed to the average face shape:

Next I took all the morphed dane faces and averaged them all to get the average dane face:

Below is my face warped into the average geometry:

Below is the average Dane face warped into my face geometry:

Caricatures: Extrapolating from the mean

Now instead of interpolating with the average face, I will make a caricature of my face by extrapolating from the average Dane face. Specifically, I will set the warp fraction to be [-1.5, -1, -0.5, 1, 1.5, 2]. If the warp fraction is negative it means that my face becomes less Dane and if it is positive then my face becomes more Dane. See the results below:

It seems like the more dane my face becomes, the more circular my face becomes.

Bells and Whistles: Change My Face's Gender

I will make my face appear more feminine. First, I found the average female face in South India:

Below is are the key points and the triangulation:

Lets first start by morphing only the shape of my face to the shape of the average Indian female:

Next I will only morph my face with the appearance of the average Indian female:

Finally I will morph both appearence and shape to get the following result: