On-line locomotion synthesis for virtual humans

Ever since the development of Computer Graphics in the industrial and academic worlds in the seventies, public knowledge and expertise have grown in a tremendous way, notably because of the increasing fascination for Computer Animation. This specific field of Computer Graphics gathers numerous techniques, especially for the animation of characters or virtual humans in movies and video games. To create such high-fidelity animations, a particular interest has been dedicated to motion capture, a technology which allows to record the 3D movement of a live performer. The resulting realism motion is convincing. However, this technique offers little control to animators, as the recorded motion can only be played back. Recently, many advances based on motion capture have been published, concerning slight but precise modifications of an original motion or the parameterization of large motion databases. The challenge consists in combining motion realism with an intuitive on-line motion control, while preserving real-time performances. In the first part of this thesis, we would like to add a brick in the wall of motion parameterization techniques based on motion capture, by introducing a generic motion modeling for locomotion and jump activities. For this purpose, we simplify the motion representation using a statistical method in order to facilitate the elaboration of an efficient parametric model. This model is structured in hierarchical levels, allowing an intuitive motion synthesis with high-level parameters. In addition, we present a space and time normalization process to adapt our model to characters of various sizes. In the second part, we integrate this motion modeling in an animation engine, thus allowing for the generation of a continuous stream of motion for virtual humans. We provide two additional tools to improve the flexibility of our engine. Based on the concept of motion anticipation, we first introduce an on-line method for detecting and enforcing foot-ground constraints. Hence, a straight line walking motion can be smoothly modified to a curved one. Secondly, we propose an approach for the automatic and coherent synthesis of transitions from locomotion to jump (and inversely) motions, by taking into account their respective properties. Finally, we consider the interaction of a virtual human with its environment. Given initial and final conditions set on the locomotion speed and foot positions, we propose a method which computes the corresponding trajectory. To illustrate this method, we propose a case study which mirrors as closely as possible the behavior of a human confronted with an obstacle: at any time, obstacles may be interactively created in front of a moving virtual human. Our method computes a trajectory allowing the virtual human to precisely jump over the obstacle in an on-line manner.

CAIAS: camera agent based on intelligent action spotting for real-time participatory animation in virtual stage

Daniel Thalmann

Nowadays, it becomes possible to show a live virtual stage application allowing a human's real time participation by motion capture. The big problem in participatory animation is figuring out where to place the camera and in what direction to point it to provide an interesting feedback. In participatory animation such as live stage performance or television show, animation is not specified in the form of giving commands but by a real human's action. Unlike filming, pre- and post-editing is not allowed. Therefore, real time decisions on the visualizing region should be made for camera placement. There have recently been several works on automatic camera control, but none of them is applicable for improvised virtual stage application because they lack the ability to find a situation in real time. The paper addresses the issues and a method for automatic determination of camera placement to visualize live virtual stage animation. Our approach is spotting an actor's action in real time by choosing a necessary level among multiple levels of perception and analysis

1999

Realtime Face Tracking and Animation

Sofien Bouaziz

Capturing and processing human geometry, appearance, and motion is at the core of computer graphics, computer vision, and human-computer interaction. The high complexity of human geometry and motion dynamics, and the high sensitivity of the human visual system to variations and subtleties in faces and bodies make the 3D acquisition and reconstruction of humans in motion a challenging task. Digital humans are often created through a combination of 3D scanning, appearance acquisition, and motion capture, leading to stunning results in recent feature films. However, these methods typically require complex acquisition systems and substantial manual post-processing. As a result, creating and animating high-quality digital avatars entails long turn-around times and substantial production costs. Recent technological advances in RGB-D devices, such as Microsoft Kinect, brought new hopes for realtime, portable, and affordable systems allowing to capture facial expressions as well as hand and body motions. RGB-D devices typically capture an image and a depth map. This permits to formulate the motion tracking problem as a 2D/3D non-rigid registration of a deformable model to the input data. We introduce a novel face tracking algorithm that combines geometry and texture registration with pre-recorded animation priors in a single optimization. This led to unprecedented face tracking quality on a low cost consumer level device. The main drawback of this approach in the context of consumer applications is the need for an offline user-specific training. Robust and efficient tracking is achieved by building an accurate 3D expression model of the user's face who is scanned in a predefined set of facial expressions. We extended this approach removing the need of a user-specific training or calibration, or any other form of manual assistance, by modeling online a 3D user-specific dynamic face model. In complement of a realtime face tracking and modeling algorithm, we developed a novel system for animation retargeting that allows learning a high-quality mapping between motion capture data and arbitrary target characters. We addressed one of the main challenges of existing example-based retargeting methods, the need for a large number of accurate training examples to define the correspondence between source and target expression spaces. We showed that this number can be significantly reduced by leveraging the information contained in unlabeled data, i.e. facial expressions in the source or target space without corresponding poses. Finally, we present a novel realtime physics-based animation technique allowing to simulate a large range of deformable materials such as fat, flesh, hair, or muscles. This approach could be used to produce more lifelike animations by enhancing the animated avatars with secondary effects. We believe that the realtime face tracking and animation pipeline presented in this thesis has the potential to inspire numerous future research in the area of computer-generated animation. Already, several ideas presented in thesis have been successfully used in industry and this work gave birth to the startup company faceshift AG.

EPFL2015

Realtime Face Tracking and Animation

Sofien Bouaziz

EPFL2015