The emerging 4D printing techniques open new horizons for fabricating self-actuated deformable objects by combing strength of 3D printing and stimuli-responsive shape memory materials. This article focuses on designing self-actuated deformable solids for 4D printing such that a solid can be programmed into a temporary shape and later recovers to its original shape after heating. To avoid a high material cost, we choose a dual-material strategy that mixes an expensive thermo-responsive shape memory polymer (SMP) material with a common elastic material, which however leads to undesired deformation at the shape programming stage. We model this shape programming process as two elastic models with different parameters linked by a median shape based on customizing a constitutive model of thermo-responsive SMPs. Taking this material modeling as a foundation, we formulate our design problem as a nonconvex optimization to find the distribution of SMP materials over the whole object as well as the median shape, and develop an efficient and parallelizable method to solve it. We show that our proposed approach is able to design self-actuated deformable objects that cannot be achieved by state of the art approaches, and demonstrate their usefulness with three example applications.

Fine-grained image classification with a few-shot classifier is a highly challenging open problem at the core of a numerous data labeling applications. In this paper, we present Few-shot Classifier Generative Adversarial Network as an approach for few-shot classification. We address the problem of few-shot classification by designing a GAN model in which the discriminator and the generator compete to output labeled data in any case. In contrast to previous methods, our techniques generate then classify images into multiple fake or real classes. A key innovation of our adversarial approach is to allow fine- grained classification using multiple fake classes with semi- supervised deep learning. A major strength of our techniques lies in its label-agnostic characteristic, in the sense that the system handles both labeled and unlabeled data during training. We validate quantitatively our few-shot classifier on the MNIST and SVHN datasets by varying the ratio of labeled data over unlabeled data in the training set. Our quantitative analysis demonstrates that our techniques produce better classification performance when using multiple fake classes and larger amount of unlabelled data.

Cage-based structures are reduced subspace deformers enabling non-isometric stretching deformations induced by clothing or muscle bulging. In this paper, we reformulate the cage-based rigging as an incompressible Stokes problem in the vorticity space. The key to our approach is a compact stencil allowing the expression of fluid-inspired high-order coordinates. Thus, our cage-based coordinates are obtained by vorticity transport as the numerical solution of the linearized Stokes equations. Then, we turn the incompressible creeping Newtonian flow into Stokes equations, and we devise a second-order compact approximation with center differencing for solving the vorticity-stream function. To the best of our knowledge, our work is the first to devise a vorticity-stream function formulation as a computational model for cage-based weighting functions.

The emergence of low-cost 3D printers steers the investigation of new geometric problems that control the quality of the fabricated object. In this paper, we present a method to reduce the material cost and weight of a given object while providing a durable printed model that is resistant to impact and external forces. We introduce a hollowing optimization algorithm based on the concept of honeycomb-cells structure. Honeycombs structures are known to be of minimal material cost while providing strength in tension. We utilize the Voronoi diagram to compute irregular honeycomb-like volume tessellations which define the inner structure. We formulate our problem as a strength--to--weight optimization and cast it as mutually finding an optimal interior tessellation and its maximal hollowing subject to relieve the interior stress. Thus, our system allows to build-to-last 3D printed objects with large control over their strength-to-weight ratio and easily model various interior structures. We demonstrate our method on a collection of 3D objects from different categories. Furthermore, we evaluate our method by printing our hollowed models and measure their stress and weights.

In this work, we describe a new and simple approach to re-use skeleton of animation with joint-based Laplacian-type regularization, in the context of exaggerated skeleton-based character animation. Despite decades of research, interactive character animations still have a lack of flexibility and editability in order to re-use real vertebral motion. In further details, generation of expressive cartoon animation from real data, is a challenging key task for non-photorealistic animation. Hence, a major problem for artists in production is to enhance the expressiveness of classical motion clips by direct manipulation of the underlying skeletal structure. Relatively small number of researchers present their approach for processing cartoon effects on motion data in [Kwon and Lee 2007; Davis and Kannappan 2002; Bregler et al. 2002]. However existing techniques often avoid dealing with the potential of skeletal-based optimization, while preserving the joint coherence and connectivity. Besides, the majority of characters in cartoons have the flexibility to stretch to extreme positions and squash to astounding shapes. It can also be noticed that squash-and-stretch is easier to realize in traditional animation rather than mocap-based computer generated animation. For this reason, Ratatouille a Pixar® movie, did not use a rigid skeleton, while abandoning motion capture to reach such essential non-ultra realistic appeal.

Recent advances in dynamic surface capture have made the creation of a realistic animation a promising task for modern visual media production such as computer generated movies or 3D cinematographic video games. Nonetheless, automatic creation of cartoon mesh animations, driven by real-life cues, is still a costly and time-consuming process and presents a number of hard technical challenges. To the best of our knowledge, we propose the first attempt at generating as-photorealistic-as possible cartoon animation from markerless surface performance capture. The purpose of synthesizing new puppetry animation demonstrating fidelity to the spirit of comic book style with more exaggerated motion while preserving extreme captured cloth wrinkles is difficult to achieve. Consequently, the key contribution of our work focuses on a novel cartoon stylization approach for 3D video that efficiently reuse temporally consistent dynamic surface sequence captured from real-world actor performance. In particular, our simple and effective algorithm converts realistic spatiotemporal captured surfaces and multi-view data into an exaggerated life-like squash-and-stretch shape evolution coupled with context-aware cartoon-style expressive rendering.