We introduce the concept of unconstrained real-time 3D facial performance capture through explicit semantic segmentation in the RGB input. To ensure robustness, cutting edge supervised learning approaches rely on large training datasets of face images captured in the wild. While impressive tracking quality has been demonstrated for faces that are largely visible, any occlusion due to hair, accessories, or hand-toface gestures would result in significant visual artifacts and loss of tracking accuracy. The modeling of occlusions has been mostly avoided due to its immense space of appearance variability. To address this curse of high dimensionality, we perform tracking in unconstrained images assuming non-face regions can be fully masked out. Along with recent breakthroughs in deep learning, we demonstrate that pixel-level facial segmentation is possible in real-time by repurposing convolutional neural networks designed originally for general semantic segmentation. We develop an efficient architecture based on a two-stream deconvolution network with complementary characteristics, and introduce carefully designed training samples and data augmentation strategies for improved segmentation accuracy and robustness. We adopt a state-of-the-art regressionbased facial tracking framework with segmented face images as training, and demonstrate accurate and uninterrupted facial performance capture in the presence of extreme occlusion and even side views. Furthermore, the resulting segmentation can be directly used to composite partial 3D face models on the input images and enable seamless facial manipulation tasks, such as virtual make-up or face replacement.

The application of auxetic composites in practice often relies on a compromise between properties as auxetics are mostly too porous (not dense enough or not stiff enough) to bear structural loads. Hence, the focus of this paper is topological design optimization of new functionally graded cellular composites with auxetics using a level set method. Firstly, a new hierarchical multi-scale formulation is developed to account for both the auxetic behaviour of the microstructure and the stiffness of the macrostructure. The composite, comprising multiple layers of periodic microstructures, is tailored to have functionally graded properties for stiffness and auxetic behaviours, subject to volumetric gradient constraints. Secondly, the microstructures underpinning composite layers are topologically designed under the consideration of boundary and loading conditions of the macrostructure. Finally, a level set method is applied to evolve the shape and topology of the microstructure for each layer, with the numerical homogenization method to evaluate the effective properties of the microstructures. Several numerical examples are used to demonstrate the effectiveness of the proposed method. It can be seen that such composites systematically gear together the features of the functionally graded materials, cellular composites, and metamaterials towards a new kind of man-made composites.

This paper proposes a new multiscale topology optimization method for the design of porous composites composed of the multi-domain material microstructures considering three design elements: the topology of the macrostructure, the topologies of multiple material microstructures and their overall distribution in the macrostructure. The multiscale design involves two optimization stages: the free material distribution optimization and the concurrent topology optimization. Firstly, the variable thickness sheet (VTS) method with the regularization mechanism is used to generate multiple element density distributions in the macro design domain. Hence, different groups of elements with the identical densities can be uniformly arranged in their corresponding domains, and each domain in the space will be periodically configured by a unique representative microstructure. Secondly, with the discrete material distributions achieved in the macro domain, the topology of the macrostructure and topologies of multiple representative microstructures are concurrently optimized by a parametric level set method combined with the numerical homogenization method. Finally. Several 2D and 3D numerical examples are provided to demonstrate the effectiveness of the proposed multiscale topology optimization method.

No abstract