Neural Network Facemorphs

face

Hübsche Spielerei mit AI-Gesichtern von Michael Flynn: Generating Faces with Deconvolution Networks. Letztlich „nur“ eine animierte Version von den bekannten Neural Network Face-Grids, wirklich spannend finde ich allerdings, was die künstliche Intelligenz aus „unmöglichen Parametern“ macht – um zum Beispiel einen Morph von einem linksgedrehten/rechtsgedrehten Profil errechnen soll, allerdings nichts über Frontal/Hinterköpfe gelernt hat, dann also Gesichts/Körperteile erfinden muss. Ergbniss: Wunderbare Morph-AI-Glitchfaces!

face

So far, all of the parameter’s we’ve given to the network have been “legal” more or less. The identity and emotion vectors have so far always been unit length (i.e: representing an even mixture of identities/emotions) and the orientations have always represented a valid angle. But what happens when we break those rules, and instead feed the network random values? It’s pretty horrific [Image above].

From some combination of not knowing how to interpret invalid orientations and having “too much” identity or emotion to process, the network begins to stretch and contort faces in really unsettling, uncanny ways. We can also create an animation out of this by shifting our inputs randomly little by little each frame: