AnimGAN: A Spatiotemporally-Conditioned Generative Adversarial Network for Character Animation
This addresses the problem of producing controlled and realistic character animations for human-AI interactions, representing an incremental improvement over traditional conditional GANs.
The paper tackled generating realistic character animations by proposing a spatiotemporally-conditioned GAN that produces sequences similar to given ones in semantics and dynamics, resulting in plausible and semantically relevant animations that match user expectations.
Producing realistic character animations is one of the essential tasks in human-AI interactions. Considered as a sequence of poses of a humanoid, the task can be considered as a sequence generation problem with spatiotemporal smoothness and realism constraints. Additionally, we wish to control the behavior of AI agents by giving them what to do and, more specifically, how to do it. We proposed a spatiotemporally-conditioned GAN that generates a sequence that is similar to a given sequence in terms of semantics and spatiotemporal dynamics. Using LSTM-based generator and graph ConvNet discriminator, this system is trained end-to-end on a large gathered dataset of gestures, expressions, and actions. Experiments showed that compared to traditional conditional GAN, our method creates plausible, realistic, and semantically relevant humanoid animation sequences that match user expectations.