Abstract
Aiming at the situation that the phase-functioned neural network can't produce natural action or interact with the world poorly in some specific scenes, this paper proposes to divide different states according to the phase information of characters, and train several independent networks with different weights to predict the next frame state of characters. The experimental results show that, while the complexity of the model is reduced, the action generated by this method is more natural and fluent, which has good application value.
Export citation and abstract BibTeX RIS
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.