Project 2

By Samuel Sunghun Lee [CS194-26 - Fall 2020]

Part 1 Defining Correspondences

We first manually define keypoints in the images so that we can triangulate. We will then triangulate with Delaunay Triangulation. The following images are produced with Delaunay Triangulation.
dx dy

Part 2 Computing the "Midway" Face

Using the Delaunay Triangles from before, we will first compute the average shape by averaging the keylocation of the two faces' keypoints. Then, we will warp both faces into each the shape we had calculated through Affine and through Polygon masking. Then, we will average the two colors together to produce an image that is the 'Midway' Face.
dx dy dy

Part 3 The Morph Sequence

From the previous section, we saw that we can compute the "Midway" Face. However, we can change the alpha values i.e the warp_alpha that we use to multiple and scale both images when we warp them into the average shape. By using this warp_alpha at many iterations, we can produce 'Alpha-way images' that act as basically the in between of images between 1 and 2. We can then stack these images together, and then we can create a GIF by scaling the alphas from 0 to 1.

Part 4 The "Mean face" of a population

We can calculate the mean face of a population by generalizing our functions from Part 1 and Part 2. I.E we can calculate the average shape of all the images from the points, use a Delaunay triangulation to calculate the average shape by doing the same thing from Part 2 but with every single image.
dy dy

Part 5 Caricatures: Extrapolating from the mean

We can create caricatures of the images from the meanshape by taking the shape and Extrapolating my face to one of the mean face images. From this, we can see that the image face became wider and the nose became larger.
We can change values of the constant alpha to get caricatures.
dy dy


This bells and whistles is the one where we morph the students images one by one. LINK TO THE YOUTUBE VIDEO FOR CLASS MORPHING MUSIC VIDEO