arXiv subjects

Andrea Tagliasacchi

Publications and source records attributed to Andrea Tagliasacchi.

97 records · Page 6Linked to original sources

arXiv2019

VoronoiNet: General Functional Approximators with Local Support

Voronoi diagrams are highly compact representations that are used in various Graphics applications. In this work, we show how to embed a differentiable version of it -- via a novel deep architecture -- into a generative deep network. By doing so, we achieve a highly compact latent embedding that is able to provide much more detailed reconstructions, both in 2D and 3D, for various shapes. In this tech report, we introduce our representation and present a set of preliminary results comparing it with recently proposed implicit occupancy networks.

cs.CV↗

arXiv2019

Linearized Multi-Sampling for Differentiable Image Transformation

We propose a novel image sampling method for differentiable image transformation in deep neural networks. The sampling schemes currently used in deep learning, such as Spatial Transformer Networks, rely on bilinear interpolation, which performs poorly under severe scale changes, and more importantly, results in poor gradient propagation. This is due to their strict reliance on direct neighbors. Instead, we propose to generate random auxiliary samples in the vicinity of each pixel in the sampled image, and create a linear approximation with their intensity values. We then use this approximation as a differentiable formula for the transformed image. We demonstrate that our approach produces more representative gradients with a wider basin of convergence for image alignment, which leads to considerable performance improvements when training networks for classification tasks. This is not only true under large downsampling, but also when there are no scale changes. We compare our approach with multi-scale sampling and show that we outperform it. We then demonstrate that our improvements to the sampler are compatible with other tangential improvements to Spatial Transformer Networks and that it further improves their performance.

cs.CV↗

arXiv2019

VIPER: Volume Invariant Position-based Elastic Rods

We extend the formulation of position-based rods to include elastic volumetric deformations. We achieve this by introducing an additional degree of freedom per vertex -- isotropic scale (and its velocity). Including scale enriches the space of possible deformations, allowing the simulation of volumetric effects, such as a reduction in cross-sectional area when a rod is stretched. We rigorously derive the continuous formulation of its elastic energy potentials, and hence its associated position-based dynamics (PBD) updates to realize this model, enabling the simulation of up to 26000 DOFs at 140 Hz in our GPU implementation. We further show how rods can provide a compact alternative to tetrahedral meshes for the representation of complex muscle deformations, as well as providing a convenient representation for collision detection. This is achieved by modeling a muscle as a bundle of rods, for which we also introduce a technique to automatically convert a muscle surface mesh into a rods-bundle. Finally, we show how rods and/or bundles can be skinned to a surface mesh to drive its deformation, resulting in an alternative to cages for real-time volumetric deformation.

cs.GR↗

arXiv2019

Volumetric Capture of Humans with a Single RGBD Camera via Semi-Parametric Learning

Volumetric (4D) performance capture is fundamental for AR/VR content generation. Whereas previous work in 4D performance capture has shown impressive results in studio settings, the technology is still far from being accessible to a typical consumer who, at best, might own a single RGBD sensor. Thus, in this work, we propose a method to synthesize free viewpoint renderings using a single RGBD camera. The key insight is to leverage previously seen "calibration" images of a given user to extrapolate what should be rendered in a novel viewpoint from the data available in the sensor. Given these past observations from multiple viewpoints, and the current RGBD image from a fixed view, we propose an end-to-end framework that fuses both these data sources to generate novel renderings of the performer. We demonstrate that the method can produce high fidelity images, and handle extreme changes in subject pose and camera viewpoints. We also show that the system generalizes to performers not seen in the training data. We run exhaustive experiments demonstrating the effectiveness of the proposed semi-parametric model (i.e. calibration images available to the neural network) compared to other state of the art machine learned solutions. Further, we compare the method with more traditional pipelines that employ multi-view capture. We show that our framework is able to achieve compelling results, with substantially less infrastructure than previously required.

cs.CV↗

arXiv2018

HandSeg: An Automatically Labeled Dataset for Hand Segmentation from Depth Images

We propose an automatic method for generating high-quality annotations for depth-based hand segmentation, and introduce a large-scale hand segmentation dataset. Existing datasets are typically limited to a single hand. By exploiting the visual cues given by an RGBD sensor and a pair of colored gloves, we automatically generate dense annotations for two hand segmentation. This lowers the cost/complexity of creating high quality datasets, and makes it easy to expand the dataset in the future. We further show that existing datasets, even with data augmentation, are not sufficient to train a hand segmentation algorithm that can distinguish two hands. Source and datasets will be made publicly available.

cs.CV↗

arXiv2018

Espresso: Efficient Forward Propagation for BCNNs

There are many applications scenarios for which the computational performance and memory footprint of the prediction phase of Deep Neural Networks (DNNs) needs to be optimized. Binary Neural Networks (BDNNs) have been shown to be an effective way of achieving this objective. In this paper, we show how Convolutional Neural Networks (CNNs) can be implemented using binary representations. Espresso is a compact, yet powerful library written in C/CUDA that features all the functionalities required for the forward propagation of CNNs, in a binary file less than 400KB, without any external dependencies. Although it is mainly designed to take advantage of massive GPU parallelism, Espresso also provides an equivalent CPU implementation for CNNs. Espresso provides special convolutional and dense layers for BCNNs, leveraging bit-packing and bit-wise computations for efficient execution. These techniques provide a speed-up of matrix-multiplication routines, and at the same time, reduce memory usage when storing parameters and activations. We experimentally show that Espresso is significantly faster than existing implementations of optimized binary neural networks ($\approx$ 2 orders of magnitude). Espresso is released under the Apache 2.0 license and is available at http://github.com/fpeder/espresso.

cs.DC↗

arXiv2014

Skeletal Representations and Applications

When representing a solid object there are alternatives to the use of traditional explicit (surface meshes) or implicit (zero crossing of implicit functions) methods. Skeletal representations encode shape information in a mixed fashion: they are composed of a set of explicit primitives, yet they are able to efficiently encode the shape's volume as well as its topology. I will discuss, in two dimensions, how symmetry can be used to reduce the dimensionality of the data (from a 2D solid to a 1D curve), and how this relates to the classical definition of skeletons by Medial Axis Transform. While the medial axis of a 2D shape is composed of a set of curves, in 3D it results in a set of sheets connected in a complex fashion. Because of this complexity, medial skeletons are difficult to use in practical applications. Curve skeletons address this problem by strictly requiring their geometry to be one dimensional, resulting in an intuitive yet powerful shape representation. In this report I will define both medial and curve skeletons and discuss their mutual relationship. I will also present several algorithms for their computation and a variety of scenarios where skeletons are employed, with a special focus on geometry processing and shape analysis.

cs.GR↗