AI Article Synopsis

  • The computation of realistic pose transitions for virtual human figures is crucial for 3D graphics and animation, but traditional methods often rely on hard-to-obtain heuristic knowledge and are limited in handling complex movements.
  • With the rise of large-scale datasets, deep learning models offer a promising alternative, though they face issues like dealing with unseen data and lack of generative capability.
  • This paper presents a hybrid approach called Dynamic Motion Transition (DMT) that combines model-driven and data-driven techniques to improve motion transition quality through data augmentation, force learning, and flexible motion creation, resulting in superior performance compared to existing methods.

Article Abstract

The rapid, accurate, and robust computation of virtual human figures' "in-between" pose transitions from available and sometimes sparse inputs is of fundamental significance to 3D interactive graphics and computer animation. Various methods have been proposed to produce natural lifelike transitions of human pose automatically in recent decades. Nevertheless, conventional pure model-driven methods require heuristic knowledge (e.g., least motion guided by physics laws) and ad-hoc clues (e.g., splines with non-uniform time warp) that are difficult to obtain, learn, and infer. With the fast emergence of large-scale datasets readily available to animators in the most recent years, deep models afford a powerful alternative to tackle the aforementioned challenges. However, pure data-driven methods still suffer from the remaining challenges such as unseen data in practice and less generative power in model/domain/data transfer, and the measurement of the generative power has always been omitted in these works. In essence, data-driven methods solely rely on the qualities and quantities of training datasets. In this paper, we propose a hybrid approach built upon the seamless integration of data-driven and model-driven methods, called Dynamic Motion Transition (DMT), with the following salient modeling advantages: (1) The data augmentation capability based on the limited human locomotion data capture and the concept of force-derived directly from physical laws; (2) Force learning by which skeleton joints are driven to move, and the Conditional Temporal Transformer (CTT) being trained to learn the force change in the local range, both at the fine level; and (3) At the coarse level, the effective and flexible creation of the subsequent step motion using Dynamic Movement Primitives (DMP) until the target is reached. Our extensive experiments have confirmed that our model can outperform the state-of-the-art methods under the newly devised metric by virtue of the least action loss function. In addition, our novel method and system are of immediate benefit to many other animation tasks such as motion synthesis and control, and motion tracking and prediction in this bigdata graphics era.

Download full-text PDF

Source
http://dx.doi.org/10.1109/TVCG.2024.3372421DOI Listing

Publication Analysis

Top Keywords

dynamic motion
8
motion transition
8
data-driven model-driven
8
human pose
8
pose transitions
8
model-driven methods
8
data-driven methods
8
generative power
8
methods
6
motion
5

Similar Publications

Article Synopsis
  • Multimodal sensing using soft body dynamics is essential for controlling soft robotic movements, particularly through the imitation of whisker dynamics.
  • The study introduces a multitasking electronic brush (e-brush) featuring four integrated pressure sensors capable of monitoring motion parameters like speed, force, slip, and surface interactions with low pressure sensitivity.
  • A reservoir computing algorithm is employed to accurately extract and analyze these motion parameters, successfully demonstrating the brush's ability to track handwriting movements as a proof of concept.
View Article and Find Full Text PDF

Proteins often harness extensive motions of domains and subunits to promote their function. Deciphering how these movements impact activity is key for understanding life's molecular machinery. The enzyme adenylate kinase is an intriguing example for this relationship; it ensures efficient catalysis by large-scale domain motions that lead to the enclosure of the bound substrates ATP and AMP.

View Article and Find Full Text PDF

Robust undulatory locomotion through neuromechanical adjustments in a dissipative medium.

J R Soc Interface

January 2025

Nantes Université, École Centrale Nantes, IMT Atlantique, CNRS, LS2N, UMR 6004, Nantes F-44000, France.

Dissipative environments are ubiquitous in nature, from microscopic swimmers in low-Reynolds-number fluids to macroscopic animals in frictional media. In this study, we consider a mathematical model of a slender elastic locomotor with an internal rhythmic neural pattern generator to examine various undulatory locomotion such as swimming and crawling behaviours. By using local mechanical load as mechanosensory feedback, we have found that undulatory locomotion robustly emerges in different rheological media.

View Article and Find Full Text PDF

Ultrashort pulses experience random quantum motion as they propagate through a mode-locked laser cavity, a phenomenon that inevitably affects the recently introduced pure-quartic solitons. Investigating this process is essential, as quantum-limited noise establishes fundamental performance limits for their application. To date, studies on quantum diffusion and the resulting timing jitter of these solitons remain sparse.

View Article and Find Full Text PDF

Creating durable, motion-compliant neural interfaces is crucial for accessing dynamic tissues under in vivo conditions and linking neural activity with behaviors. Utilizing the self-alignment of nano-fillers in a polymeric matrix under repetitive tension, here, we introduce conductive carbon nanotubes with high aspect ratios into semi-crystalline polyvinyl alcohol hydrogels, and create electrically anisotropic percolation pathways through cyclic stretching. The resulting anisotropic hydrogel fibers (diameter of 187 ± 13 µm) exhibit fatigue resistance (up to 20,000 cycles at 20% strain) with a stretchability of 64.

View Article and Find Full Text PDF

Want AI Summaries of new PubMed Abstracts delivered to your In-box?

Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!