Here, we build a system for morphing between faces (or in theory any images with similar structure). We do so by defining corresponding feature points for each of the two input images, computing a triangulation over those points, interpolating between the feature points to create an output “morphed” shape, and finally warping each image into that shape.
Using this program, we can easily create fun morphing sequences by varying our interpolation parameter over time. For example, when relatives claim our dog looks like Bill Cosby, we can [not use our program to] create a side-by-side comparison and [actually use our program to create] an apparently entertaining GIF.
We’re also able to “alter people’s ethnicity” by morphing a subject into the average face of any given country. We do so below with my old roommate Oliver and your average Japanese actor (truly).