For this project, I chose to morph my face with that of our lovely chancellor, Carol T. Christ. I computed the triangulations by manually selecting a sequence of points in the same order on both my and Carol's faces. The points I decided to select are according to the image below. Then, I calculated the Delaunay triangulation at the midway shape (i.e., average of me and Carol's points) to lessen the potential triangle deformations. To get the morphing sequence, at each timestep t (in [0, 1]), I calculated the key points of the warped shape as (1 - t) * jie_pts + t * carol_pts. Then, I looped through all the triangles of this warped shape, and for each of them, I mapped the pixels in them to their corresponding locations in the original images. After that, I set the pixel values to a weighted average of the corresponding pixels in the original images.
For this part, I used the Dane's dataset and computed the average face of all neutral-expression Danish females. Below is the average face.
The key points of each picture are also given in the dataset. Using the techniques in the previous part (first calculating the warped shape then fill in the right pixels), I was able to successfully warp my face into the average geometry, and vice versa. The results are shown below.
1. I morphed my face into an average Danish male to change my gender. Results are shown below.
The midway pictures are probably good representations of me as a male. I have included them below.
2. I made a music video on morphing EECS faculty's faces. I chose the background music to be a well-known cal-band song. Here is the youtube link
3. Here is the face-morphing music video of the students in the class!