File Download
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.5244/c.31.119
- Scopus: eid_2-s2.0-85088774383
Supplementary
-
Citations:
- Scopus: 0
- Appears in Collections:
Conference Paper: A recurrent variational autoencoder for human motion synthesis
Title | A recurrent variational autoencoder for human motion synthesis |
---|---|
Authors | |
Issue Date | 2017 |
Citation | 28th British Machine Vision Conference (BMVC 2017), London, 4-7 September 2017. In Proceedings of the British Machine Vision Conference (BMVC), 2017, p. 119.1-119.12 How to Cite? |
Abstract | We propose a novel generative model of human motion that can be trained using a large motion capture dataset, and allows users to produce animations from high-level control signals. As previous architectures struggle to predict motions far into the future due to the inherent ambiguity, we argue that a user-provided control signal is desirable for animators and greatly reduces the predictive error for long sequences. Thus, we formulate a framework which explicitly introduces an encoding of control signals into a variational inference framework trained to learn the manifold of human motion. As part of this framework, we formulate a prior on the latent space, which allows us to generate high-quality motion without providing frames from an existing sequence. We further model the sequential nature of the task by combining samples from a variational approximation to the intractable posterior with the control signal through a recurrent neural network (RNN) that synthesizes the motion. We show that our system can predict the movements of the human body over long horizons more accurately than state-of-the-art methods. Finally, the design of our system considers practical use cases and thus provides a competitive approach to motion synthesis. |
Persistent Identifier | http://hdl.handle.net/10722/288827 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Habibie, Ikhsanul | - |
dc.contributor.author | Holden, Daniel | - |
dc.contributor.author | Schwarz, Jonathan | - |
dc.contributor.author | Yearsley, Joe | - |
dc.contributor.author | Komura, Taku | - |
dc.date.accessioned | 2020-10-12T08:05:59Z | - |
dc.date.available | 2020-10-12T08:05:59Z | - |
dc.date.issued | 2017 | - |
dc.identifier.citation | 28th British Machine Vision Conference (BMVC 2017), London, 4-7 September 2017. In Proceedings of the British Machine Vision Conference (BMVC), 2017, p. 119.1-119.12 | - |
dc.identifier.uri | http://hdl.handle.net/10722/288827 | - |
dc.description.abstract | We propose a novel generative model of human motion that can be trained using a large motion capture dataset, and allows users to produce animations from high-level control signals. As previous architectures struggle to predict motions far into the future due to the inherent ambiguity, we argue that a user-provided control signal is desirable for animators and greatly reduces the predictive error for long sequences. Thus, we formulate a framework which explicitly introduces an encoding of control signals into a variational inference framework trained to learn the manifold of human motion. As part of this framework, we formulate a prior on the latent space, which allows us to generate high-quality motion without providing frames from an existing sequence. We further model the sequential nature of the task by combining samples from a variational approximation to the intractable posterior with the control signal through a recurrent neural network (RNN) that synthesizes the motion. We show that our system can predict the movements of the human body over long horizons more accurately than state-of-the-art methods. Finally, the design of our system considers practical use cases and thus provides a competitive approach to motion synthesis. | - |
dc.language | eng | - |
dc.relation.ispartof | Proceedings of the British Machine Vision Conference (BMVC) | - |
dc.rights | © 2017. The copyright of this document resides with its authors. It may be distributed unchanged freely in print or electronic forms. | - |
dc.title | A recurrent variational autoencoder for human motion synthesis | - |
dc.type | Conference_Paper | - |
dc.description.nature | published_or_final_version | - |
dc.identifier.doi | 10.5244/c.31.119 | - |
dc.identifier.scopus | eid_2-s2.0-85088774383 | - |
dc.identifier.spage | 119.1 | - |
dc.identifier.epage | 119.12 | - |