Project 3: Face Morphing

In this section, I select corresponding points between two faces. These points define key facial features such as the outline of eyes, nose, mouth, etc. allowing us to establish a triangulation for morphing. In order do this, I used the tool linked on the project spec. The points shown below are what serve as a basis for computing a triangulation mesh using Delaunay Triangulation, I then use the triangulated correspondance points to comute a warp which is carried out in the next part. the face geometrically.

In this section, I use the correspondance points from Part 1, to compute the average geometry between two faces by averaging their correspondances I then compute the affine transformation matrix to go from the triangulated geometry of the mid-way face to the original images. These matricies are then used to morph each of the faces into this mid-way shape. Then I carry out interpolation in order to set the color of the pixels in the images..

In this section, I generate a sequence of frames that gradually morphs one face into the other. The transition is controlled by linearly warping the geometry and cross dissolving the colors of the pixels, creating a smooth animation between the two faces. To accomplish this, I added a warp_factor and dissolve_factor parameter to my function that generates the mid-way face. Then, I created a range of these values from [0, 1] with 45 steps in order to create a transition with 45 frames. I then compiled all of these 45 frames into a gif to display the morphing sequenece.

In this section, I compute the average face of a population retreived from the dataset linked on the project spec (FEI face database). To accomplish this, I first average all of the correspondance points of all the images in the dataset in order to get the average shape. I then create a triangulation of this shape and warp all of the faces in the dataset to this geometry. I average all of the outputs at the end to get the "mean face" of the population at the end. by morphing several faces into the mean geometry of the group. I then define new correspondances on this mean face and my face in order to warp my face into that of the average of the population.

In this section, I create caricatures by extrapolating from the mean face. In order to do this, I utilized the same morphing function that I had, but I set the warp_factor = -0.5 to get more of my face's geometry and 1.5 to get more of the average face's geometry.

In this section, I attempt to change the gender of my face by utilizing an image of the average Dutch woman. I define correspondances between this image and my image and then warp my face's geometry into the geometry of the average Dutch woman. I also warp the geometry of the average Dutch woman into my geomtry as well, and then morph both the shape and appearance.