Artificial intelligences are incredibly great at creating faces and recognizing patterns with them. I wanted to see if a computer could replicate my face by using as little information about the facial expression I was making.

To do this I programmed my own machine learning model called an autoencoder. This may be too complicated to explain here but below is a video of the computer learning to make my face after hours and hours of training. What you’ll notice about the video is that as time goes on, the computer becomes much more animated and exaggerated with it's facial movements. [Note: None of these images are real!! They are all computer generated.] Feel free to pause at any point to look at the image for yourself!

Recenty, I revisited this project using more modern teqniques to improve the results. This time I manipulated the network such that I could inject my own values into it. This allows for the possibility to customize my face. Below is a video of some successful results I got [Note: None of these images are real!! They are all computer generated.]:

For another test, I got an even larger dataset, except this time of my mouth alone, from a previous project I did to lip sync an animated character and trained it on that. Here is some images produced [Note: None of these images are real!! They are all computer generated.]: