Project 3: Face Morphing

Part 1: Defining Correspondences

I used the pictures of George Clooney and Liam Hemsworth for this part. I used ginput to get the keypoints for each of the images and then used Delaunay triangulation on the average.

Keypoints (45 points for each)

Average of key points

Delaunay

Part 2: Computing The "Mid-way Face"

In order to find the mid-way face, I find the average shape first using part 1 points and then use the affine transformation to warp George and Liam to the average shape and finally, average the colors.

Part 3: The Morph Sequence

In order to get the morph sequence, I did the same process as part 2 but used different weights for each of the images.

https://youtu.be/XADHsUypyVM

The "Mean Face" Of A Population

I downloaded the Danes/IMM dataset to find the mean face of men and women that were in the population and using a similar method as part 2 to find the mean face. Since there weren't a whole lot of women data, the mean face is not as smooth as that of men's

For the average mean face, I first find the average key points, then sum all the morphed images and divide that by the number of images. My average mean face does not come out too well, probably because of a mistake in my wrapping function.

Conclusion

Although I could not complete the project, I had a great time understanding how to make a morph sequence and really enjoyed the process of doing so.