Project 4 Submission

Name: Rohan Murthy(cs194-26-adk)


In this project I used geometric transformations to mix my face with the face of George Clooney! Essentially the way I did this was create face points for my face and George's. I used these points to find the average triangulation. Using this average triangulation I went through all points of the resulting image, figured out the transformation from the average triangulation triangle that each point is in and got the transformation matrix from that point to the corresponding point in my face and in the George face images. Thus, I averaged the pixel values from these corresponding points to the point in the result image.

Me to George Clooney

This is the triangulation point structure I used:

This is the original picture of me, the original picture of George, and the mid-way face!

This is a gif from my face to George's!

Mean Face of Population

For this section, I used the Danes database and found the population average of faces from the 40 people that were included. This database had multiple faces per person, so I just used the one that was described on the website as "Full frontal face, neutral expression, diffuse light".
This is how the points were set up for these faces:

PLEASE NOTE THE POINTS SHOWN ABOVE: Throughout this section, it may appear that pictures look a little deformed, but these deformed parts are places that ARE NOT in the facial area.
This is the Mean Face of the population that I got!

Here are some examples of a couple people in the population morphed into this average shape:
Original person 1 and person 1 morphed into average face of population:

Original person 2 and person 2 morphed into average face of population:

These morphed faces make sense, because these two people have fairly skinny faces, while the average face shape is rather wide.
This is my face!

This is my face warped into the this average face geometry!

This is the population average face warped into my face geometry!

My face is definitely wider than the population's average face. Thus, it makes sense that this facial feature is toned down in my warped face and this facial feature is accentuated in the warped average population face. Also my crooked nose feature and fuller lips are shown in the warped average face.


So my facial features include a wide face and crooked nose. I accentuated these features by finding the difference between my facial points and the population average points and added them to my points with an alpha weight.
Here is my face morphed to the average population face geometry from above:

Here are caricatures of my face with alpha 1(adding the entire difference back), and alpha 0.5(adding half the difference back):

In my caricature, my nose definitely looks more crooked and my face definitely looks wider.

Bells and Whistles: Average Indian Male

I am Indian, but how much does my face differ from that of the average Indian male?
This is me and a picture of the average Indian male:

For the triangulations, I used the same point structure from the first section (morphing into George Clooney).
This is my face morphed into the shape, appearance, and both shape and appearance of the average Indian male!