Sofie Yang
CS 194-26 - Project 3 Face Morphing
Project Overview
This project explored using affine transformations and linear interpolation to create face morphs and caricatures.
Part 1: Defining Correspondences
Sofie Delaunay Triangulation
Sofie Delaunay Triangulation Applied to Taylor Swift Points
Part 2: Computing the "Mid-way Face"
Sofie
Sofie Taylor Swift Midway Face
Taylor Swift
Part 3: The Morph Sequence"
Morph Sequence Between Sofie and Taylor Swift
Part 3: The "Mean Face" of a Population"
I used the FEI database (https://fei.edu.br/~cet/facedatabase.html) spatially aligned images to get 200 images of both male and female unsmiling (a) subjects. You can see that a person with a thin long face got a rounder face after morphing and the person with a square face and thin lips got a longer face and thicker lips to become more "average".
Average Brazillian Face
Subject 2a
Subject 2a Morphed to Average
Subject 11a
Subject 11a Morphed to Average
Sofie Aligned
Average Brazillian Face
Sofie Morphed to Average Shape
Average Morphed to Sofie Shape
The result on my face has some distortions on the forehead because the points labeled did not include any on the forehead, so that area was not affine transformed like the rest. There is also an artifact on Average Morphed to Sofie Shape because the Delaunay triangles constructed from the points on my face ended up closely overlapping when applied to the average face.
Part 4: Caricatures"
I computed my caricature off of the average Brazillian face. The features that it exaggerated are my straight eyebrows, drooping eyes and small mouth/nose. Interestingly, it also picked up on the fact that my face is not perfectly symmetrical and expanded the left larger side while reducing the right size.
Sofie Aligned
Sofie Caricature alpha = 2
Sofie Caricature alpha = 2.5
Sofie Caricature alpha = 3
Part 5: Bells and Whistles"
A few friends in this class decided to create a chain of face morphs to each other. Along the way, we had some good laughs at the similarity or distance between our faces. Here is the Youtube video link if the embedding doesn't work (https://www.youtube.com/watch?v=sTT0-nN3vtc&feature=youtu.be).
Most interesting thing from this assignment
My favorite part of this project was figuring out how affine transformations could be calculated and applied to each triangle. It was also interesting to see, computationally, the unique features of my/my friends' faces that are brought out during averaging or caricaturing.