Figure 1. HoloGAN learns to separate pose from identity (shape and appearance) only from unlabelled 2D images without sacrificing the visual fidelity of the generated images. All results shown here are sampled from HoloGAN for the same identities in each row but in different poses. AbstractWe propose a novel generative adversarial network (GAN) for the task of unsupervised learning of 3D representations from natural images. Most generative models rely on 2D kernels to generate images and make few assumptions about the 3D world. These models therefore tend to create blurry images or artefacts in tasks that require a strong 3D understanding, such as novel-view synthesis. HoloGAN instead learns a 3D representation of the world, and to render this representation in a realistic manner. Unlike other GANs, HoloGAN provides explicit control over the pose of generated objects through rigid-body transformations of the learnt 3D features. Our experiments show that using explicit 3D features enables HoloGAN to disentangle 3D pose and identity, which is further decomposed into shape and appearance, while still being able to generate images with similar or higher visual quality than other generative models. HoloGAN can be trained end-to-end from unlabelled 2D images only. In particular, we do not require pose labels, 3D shapes, or multiple views of the same objects. This shows that HoloGAN is the first generative model that learns 3D representations from natural images in an entirely unsupervised manner.

In architectural freeform design, the relation between shape and fabrication poses new challenges and requires more sophistication from the underlying geometry. The new concept of conical meshes satisfies central requirements for this application: They are quadrilateral meshes with planar faces, and therefore particularly suitable for the design of freeform glass structures. Moreover, they possess a natural offsetting operation and provide a support structure orthogonal to the mesh. Being a discrete analogue of the network of principal curvature lines, they represent fundamental shape characteristics. We show how to optimize a quad mesh such that its faces become planar, or the mesh becomes even conical. Combining this perturbation with subdivision yields a powerful new modeling tool for all types of quad meshes with planar faces, making subdivision attractive for architecture design and providing an elegant way of modeling developable surfaces.

No abstract

No abstract

The realization of quantum spin Hall effect in HgTe quantum wells is considered a milestone in the discovery of topological insulators. Quantum spin Hall states are predicted to allow current flow at the edges of an insulating bulk, as demonstrated in various experiments. A key prediction yet to be experimentally verified is the breakdown of the edge conduction under broken time-reversal symmetry. Here we first establish a systematic framework for the magnetic field dependence of electrostatically gated quantum spin Hall devices. We then study edge conduction of an inverted quantum well device under broken time-reversal symmetry using microwave impedance microscopy, and compare our findings to a non-inverted device. At zero magnetic field, only the inverted device shows clear edge conduction in its local conductivity profile, consistent with theory. Surprisingly, the edge conduction persists up to 9 T with little change. This indicates physics beyond simple quantum spin Hall model, including material-specific properties and possibly many-body effects.

The computation of a rigid body transformation which optimally aligns a set of measurement points with a surface and related registration problems are studied from the viewpoint of geometry and optimization. We provide a convergence analysis for widely used registration algorithms such as ICP, using either closest points (Besl and McKay, 1992) or tangent planes at closest points (Chen and Medioni, 1991) and for a recently developed approach based on quadratic approximants of the squared distance function . ICP based on closest points exhibits local linear convergence only. Its counterpart which minimizes squared distances to the tangent planes at closest points is a Gauss-Newton iteration; it achieves local quadratic convergence for a zero residual problem and-if enhanced by regularization and step size control-comes close to quadratic convergence in many realistic scenarios. Quadratically convergent algorithms are based on the approach in . The theoretical results are supported by a number of experiments; there, we also compare the algorithms with respect to global convergence behavior, stability and running time.

No abstract

No abstract