Active Learning for Interactive Audio-Animatronic Performance Design

Castellon, Joel; B�cher, Moritz; McCrory, Matt; Ayala, Alfredo; Stolarz, Jeremy; Mitchell, Kenny

Abstract

We present a practical neural computational approach for interactive design of Audio-Animatronic® facial performances. An offline quasi-static reference simulation, driven by a coupled mechanical assembly, accurately predicts hyperelastic skin deformations. To achieve interactive digital pose design, we train a shallow, fully connected neural network (KSNN) on input motor activations to solve the simulated mesh vertex positions. Our fully automatic synthetic training algorithm enables a first-of-its-kind learning active learning framework (GEN-LAL) for generative modeling of facial pose simulations. With adaptive selection, we significantly reduce training time to within half that of the unmodified training approach for each new Audio-Animatronic® figure.

Citation

Castellon, J., Bächer, M., McCrory, M., Ayala, A., Stolarz, J., & Mitchell, K. (2020). Active Learning for Interactive Audio-Animatronic Performance Design. The Journal of Computer Graphics Techniques, 9(3), 1-19

• Journal Article Type: Article
• Acceptance Date: Mar 12, 2020
• Online Publication Date: Oct 11, 2020
• Publication Date: Oct 11, 2020
• Deposit Date: Oct 16, 2020
• Publicly Available Date: Oct 16, 2020
• Journal: Journal of Computer Graphics Techniques
• Print ISSN: 2331-7418
• Peer Reviewed: Peer Reviewed
• Volume: 9
• Issue: 3
• Pages: 1-19
• Keywords: Deep Learning
• Public URL: http://researchrepository.napier.ac.uk/Output/2693867
• Publisher URL: http://jcgt.org/published/0009/03/01/