Motion Style Retargeting to Characters With Different Morphologies

Abstract

We present a novel approach for style retargeting to non-humanoid characters by allowing extracted stylistic features from one character to be added to the motion of another character with a different body morphology. We introduce the concept of groups of body parts (GBPs), for example, the torso, legs and tail, and we argue that they can be used to capture the individual style of a character motion. By separating GBPs from a character, the user can define mappings between characters with different morphologies. We automatically extract the motion of each GBP from the source, map it to the target and then use a constrained optimization to adjust all joints in each GBP in the target to preserve the original motion while expressing the style of the source. We show results on characters that present different morphologies to the source motion from which the style is extracted. The style transfer is intuitive and provides a high level of control. For most of the examples in this paper, the definition of GBP takes around 5 min and the optimization about 7 min on average. For the most complicated examples, the definition of three GBPsand their mapping takes about 10 min and the optimization another 30 min.

Survey on Style in 3D Human Body Motion: Taxonomy, Data, Recognition and Its Applications

“…Still, common characteristics appear such as emotions, gender, age, physical states, personality features and behaviors, that can thus be considered as subsets of style. It is in accordance with the definitions of style from Troje [29] and Abdul-Massih et al [28], who respectively claimed that style referred to "emotions, personality or biological features, such as age or gender" and "your perceived mood, behaviour or physical properties of the motion". It was confirmed by Etemad and Arya [14] who specified that the style, represented by secondary themes, contained "variations caused by individual characteristics of the actor such as gender, age, emotions, energy, mood, health, and even inherited characteristics".…”

“…On the contrary, motion decomposition was much more frequent in style transfer studies. It could be spatially [28] but it was mostly temporally [1], [8], [17].…”

“…[1] -and their emotional transforms, Abdul-Massih et al [28] or Yumer and Mitra [8] -in the frequency domain, or even Etemad and Arya [14] with neural networks. It could also be the difference between two stylistic motions such as for Xia et al [7], where the relationship between input and output styles was explained by a time-varying autoregressive model whose parameters were estimated via a linear least square regression, or for Hsu et al [2].…”

“…Motion alignment has also been widely used in motion style transfer studies. Abdul-Massih et al [28] decided to achieve it manually when groups of body parts were set by users. DTW [17] or its variants -ITW [2], CoTW [14] -was often used.…”

“…Abdul-Massih et al [28] decided to manually split spatially the skeletons into groups of body parts. They extracted positional and angular amplitude features.…”