User‐Guided Facial Animation through an Evolutionary Interface

Abstract

We propose a design framework to assist with user‐generated content in facial animation — without requiring any animation experience or ground truth reference. Where conventional prototyping methods rely on handcrafting by experienced animators, our approach looks to encode the role of the animator as an Evolutionary Algorithm acting on animation controls, driven by visual feedback from a user. Presented as a simple interface, users sample control combinations and select favourable results to influence later sampling. Over multiple iterations of disregarding unfavourable control values, parameters converge towards the user's ideal. We demonstrate our framework through two non‐trivial applications: creating highly nuanced expressions by evolving control values of a face rig and non‐linear motion through evolving control point positions of animation curves.

Using genetic algorithms to uncover individual differences in how humans represent facial emotion

“…The combination of six generations and 10 faces per generation was chosen to strike a balance between avoiding information overload for participants on each trial, increasing efficiency, presenting an adequate number of generations [38], minimizing participant 'burn-out' (i.e. boredom) since the faces become more similar with each iteration, and keeping the faces easy to view on a single screen [35]. To compensate for this relatively small population of test faces, we employed a coarse-to-fine approach to more extensively cover the search space.…”

“…A comprehensive description of the GA is presented in the electronic supplementary material, eAppendix-1 and in [35]. Expressions were generated from an animator's face rig [36] (https://www.…”

“…To apply corrective shapes to improve realism of the faces generated, we sampled from rig controls, which incorporates correctives. This procedure is described at length in [35]. Obtaining a suitable combination of blendshapes in the faces presented to participants also necessitates incorporating an appropriate mutation rate in the GA, described in terms of the difference in blendshape weights bounded [0, 1].…”

“…Obtaining a suitable combination of blendshapes in the faces presented to participants also necessitates incorporating an appropriate mutation rate in the GA, described in terms of the difference in blendshape weights bounded [0, 1]. To achieve this, a random value was assigned between 0.25 and 0.75 [35] (for complete details, see electronic supplementary material, eAppendix-1).…”

EmoGen: Quantifiable Emotion Generation and Analysis for Experimental Psychology

“…In Section II, we sketch the fields of facial dynamics modelling and genetic algorithm applications. We focus on comparing the proposed EmoGen framework to its prototype variant [2], highlighting the significant improvements in the static expression generation with respect to robustness, freedom of model space exploration and visual sample quality, as well as EmoGen's substantive extension in terms of data analysis tools and performance assessment. Then, in Section III, we discuss the principles of blendshape modelling and mesh correction automation, the chosen GA-approach and its application to facial modelling.…”