Motion Reconstruction from Variable Sparse Tracking Signals via Latent Space Optimization

View the PDF file from the paper entitled Dragposer: Reconstruction of the movement from changing changing tracking signals by improving the inherent space, written by Jose Luis Bonton and 4 other authors

PDF HTML (experimental) view

a summary:High quality rebuilding that follows user movements with advanced MOCAP systems can be achieved with many sensors. However, getting this animation quality with fewer input devices is gaining popularity because they provide MoCAP closer to the general public. The main challenges include the loss of the final response accuracy in the learning -based approach, or the absence of natural and smoothness in IK solutions. In addition, these systems are often seized carefully with a specific number of followers, which are very sensitive to lost data, for example, in scenarios where the sensor or breakdowns are dismantled. In response to these challenges, we offer Dragposer, a new system for rebuilding deep -learning movement that is accurately represented by solid and dynamic restrictions based on accuracy, as it has accurately in setting high -end effects in actual time. This is achieved through the process of improving the formation within an organized underlying space. Our system requires only one time training on the large human movement data group, and then the restrictions can be defined as losses as losses, while the situation is repeatedly repeated by calculating the gradients of these losses within the inherent area. To increase the strengthening of our approach, we merge a time prediction network, which uses a transformer structure for the coding of time directly inside the underlying area. This network guarantees that realistic improvement is limited to the difficult situation, and it also benefits from the previous data to predict a temporary coherent presentation. The results show that Dragposer outperforms both the IK methods and the latest methods that rely on data to achieve exactly final reserved locations, while resulting in natural conditions and temporary coherent movement. In addition, our system displays durability in exchange for adjustments to restrictions to flying, and shows the ability to adapt exceptional with many input configurations and changes.