Showing 1–8 of 8 results for author: Paulin, M

Harmonics Virtual Lights : fast projection of luminance field on spherical harmonics for efficient rendering

Authors: Pierre Mézières, François Desrichard, David Vanderhaeghe, Mathias Paulin

Abstract: In this paper, we introduce Harmonics Virtual Lights (HVL), to model indirect light sources for interactive global illumination of dynamic 3D scenes. Virtual Point Lights (VPL) are an efficient approach to define indirect light sources and to evaluate the resulting indirect lighting. Nonetheless, VPL suffer from disturbing artifacts, especially with high frequency materials. Virtual Spherical Ligh… ▽ More In this paper, we introduce Harmonics Virtual Lights (HVL), to model indirect light sources for interactive global illumination of dynamic 3D scenes. Virtual Point Lights (VPL) are an efficient approach to define indirect light sources and to evaluate the resulting indirect lighting. Nonetheless, VPL suffer from disturbing artifacts, especially with high frequency materials. Virtual Spherical Lights (VSL) avoid these artifacts by considering spheres instead of points but estimates the lighting integral using Monte Carlo which results to noise in the final image. We define HVL as an extension of VSL in a Spherical Harmonics (SH) framework, defining a closed form of the lighting integral evaluation. We propose an efficient SH projection of spherical lights contribution faster than existing methods. Computing the outgoing luminance requires $\mathcal{O}(n)$ operations when using materials with circular symmetric lobes, and $\mathcal{O}(n^2)$ operations for the general case, where $n$ is the number of SH bands. HVL can be used with either parametric or measured BRDF without extra cost and offers control over rendering time and image quality, by either decreasing or increasing the band limit used for SH projection. Our approach is particularly well designed to render medium-frequency one-bounce global illumination with arbitrary BRDF in interactive time. △ Less

Submitted 5 January, 2022; originally announced January 2022.

Comments: 17 pages (4 more in supplemental), under revision by the journal "Computer Graphics Forum"

ACM Class: I.3.7; I.3.6

PCEDNet : A Lightweight Neural Network for Fast and Interactive Edge Detection in 3D Point Clouds

Authors: Chems-Eddine Himeur, Thibault Lejemble, Thomas Pellegrini, Mathias Paulin, Loic Barthe, Nicolas Mellado

Abstract: In recent years, Convolutional Neural Networks (CNN) have proven to be efficient analysis tools for processing point clouds, e.g., for reconstruction, segmentation and classification. In this paper, we focus on the classification of edges in point clouds, where both edges and their surrounding are described. We propose a new parameterization adding to each point a set of differential information o… ▽ More In recent years, Convolutional Neural Networks (CNN) have proven to be efficient analysis tools for processing point clouds, e.g., for reconstruction, segmentation and classification. In this paper, we focus on the classification of edges in point clouds, where both edges and their surrounding are described. We propose a new parameterization adding to each point a set of differential information on its surrounding shape reconstructed at different scales. These parameters, stored in a Scale-Space Matrix (SSM), provide a well suited information from which an adequate neural network can learn the description of edges and use it to efficiently detect them in acquired point clouds. After successfully applying a multi-scale CNN on SSMs for the efficient classification of edges and their neighborhood, we propose a new lightweight neural network architecture outperforming the CNN in learning time, processing time and classification capabilities. Our architecture is compact, requires small learning sets, is very fast to train and classifies millions of points in seconds. △ Less

Submitted 28 September, 2021; v1 submitted 3 November, 2020; originally announced November 2020.

Journal ref: ACM Transactions on Graphics, Volume 41, Issue 1, 2022

Magnetic properties of cobalt doped ZrO$_2$ nanoparticles: Evidence of Co segregation

Authors: Francisco González Pinto, Mariano Andrés Paulin, Ana Gabriela Leyva, Joaquín Sacanell

Abstract: We synthesized pure and Co-doped (6.25 12.5 at.) ZrO$_2$ nanopowders in order to study their magnetic properties.We analyzed magnetic behavior as a function of the amount of Co and the oxygenation, which was controlled by low pressure thermal treatments. As prepared pure and Co-doped samples are diamagnetic and paramagnetic respectively. Ferromagnetism can be induced by performing low pressure the… ▽ More We synthesized pure and Co-doped (6.25 12.5 at.) ZrO$_2$ nanopowders in order to study their magnetic properties.We analyzed magnetic behavior as a function of the amount of Co and the oxygenation, which was controlled by low pressure thermal treatments. As prepared pure and Co-doped samples are diamagnetic and paramagnetic respectively. Ferromagnetism can be induced by performing low pressure thermal treatments, which becomes stronger as the dwell time of the thermal treatment is increased. This behavior can be reversed, recovering the initial diamagnetic or paramagnetic behavior, by performing reoxidizing thermal treatments. Also, a cumulative increase can be observed in the saturation of the magnetization with the number of low pressure thermal treatments performed. We believe that this phenomenon indicates that cobalt segregation induced by the thermal treatments is the responsible for the magnetic properties of the ZrO$_2$ Co system. △ Less

Submitted 18 June, 2018; originally announced June 2018.

Journal ref: Materials Research Express 5 (2018) 066103

Convolutional Patch Representations for Image Retrieval: an Unsupervised Approach

Authors: Mattis Paulin, Julien Mairal, Matthijs Douze, Zaid Harchaoui, Florent Perronnin, Cordelia Schmid

Abstract: Convolutional neural networks (CNNs) have recently received a lot of attention due to their ability to model local stationary structures in natural images in a multi-scale fashion, when learning all model parameters with supervision. While excellent performance was achieved for image classification when large amounts of labeled visual data are available, their success for un-supervised tasks such… ▽ More Convolutional neural networks (CNNs) have recently received a lot of attention due to their ability to model local stationary structures in natural images in a multi-scale fashion, when learning all model parameters with supervision. While excellent performance was achieved for image classification when large amounts of labeled visual data are available, their success for un-supervised tasks such as image retrieval has been moderate so far. Our paper focuses on this latter setting and explores several methods for learning patch descriptors without supervision with application to matching and instance-level retrieval. To that effect, we propose a new family of convolutional descriptors for patch representation , based on the recently introduced convolutional kernel networks. We show that our descriptor, named Patch-CKN, performs better than SIFT as well as other convolutional networks learned by artificially introducing supervision and is significantly faster to train. To demonstrate its effectiveness, we perform an extensive evaluation on standard benchmarks for patch and image retrieval where we obtain state-of-the-art results. We also introduce a new dataset called RomePatches, which allows to simultaneously study descriptor performance for patch and image retrieval. △ Less

Submitted 1 March, 2016; originally announced March 2016.

The Origin of Inference: Ediacaran Ecology and the Evolution of Bayesian Brains

Authors: Michael G. Paulin

Abstract: The evolution of spiking neurons and nervous systems in the late Ediacaran period simultaneously with the evolution of carnivores around 550 million years ago can be explained by the need for accurately timed decisions under an imminent threat of being eaten. A simple model shows that threshold triggering devices, spiking neurons, are utility-maximizing decision-makers for the timing of escape ref… ▽ More The evolution of spiking neurons and nervous systems in the late Ediacaran period simultaneously with the evolution of carnivores around 550 million years ago can be explained by the need for accurately timed decisions under an imminent threat of being eaten. A simple model shows that threshold triggering devices, spiking neurons, are utility-maximizing decision-makers for the timing of escape reflexes given the sensory cues available to Ediacaran animals at the onset of carnivory. Decisions are suboptimal for very weak stimuli, providing selection pressure for secondary processing of primary spike train data. A simple network can make approximately Bayes optimal decisions given stochastic spike trains. Decisions that are arbitrarily close to Bayes optimal can be obtained by enlarging this network. A subnetwork that computes the Bayesian posterior density of the critical state variable - distance between predator and prey - emerges as a core component of the decision-making mechanism. This is a neural analog of a Bayesian particle filter with cerebellar-like architecture. The model explains fundamental properties of neurons and nervous systems in modern animals and makes testable predictions. △ Less

Submitted 11 April, 2015; originally announced April 2015.

MSC Class: 92C20

Bayesian Inference with Spiking Neurons

Authors: Michael G. Paulin, Andre van Schaik

Abstract: Humans and other animals behave as if we perform fast Bayesian inference underlying decisions and movement control given uncertain sense data. Here we show that a biophysically realistic model of the subthreshold membrane potential of a single neuron can exactly compute the numerator in Bayes rule for inferring the Poisson parameter of a sensory spike train. A simple network of spiking neurons can… ▽ More Humans and other animals behave as if we perform fast Bayesian inference underlying decisions and movement control given uncertain sense data. Here we show that a biophysically realistic model of the subthreshold membrane potential of a single neuron can exactly compute the numerator in Bayes rule for inferring the Poisson parameter of a sensory spike train. A simple network of spiking neurons can construct and represent the Bayesian posterior density of a parameter of an external cause that affects the Poisson parameter, accurately and in real time. △ Less

Submitted 19 June, 2014; originally announced June 2014.

Reversible switching of room temperature ferromagnetism in CeO2-Co nanoparticles

Authors: J. Sacanell, M. A. Paulin, V. Ferrari, G. Garbarino, A. G. Leyva

Abstract: We investigated the reversible ferromagnetic (FM) behavior of pure and Co doped CeO2 nanopowders. The as-sintered samples displayed an increasing paramagnetic contribution upon Co do**. Room temperature FM is obtained simply by performing thermal treatments in vacuum at temperatures as low as 500^{\circ}C and it can be switched off by performing thermal treatments in oxidizing conditions. The FM… ▽ More We investigated the reversible ferromagnetic (FM) behavior of pure and Co doped CeO2 nanopowders. The as-sintered samples displayed an increasing paramagnetic contribution upon Co do**. Room temperature FM is obtained simply by performing thermal treatments in vacuum at temperatures as low as 500^{\circ}C and it can be switched off by performing thermal treatments in oxidizing conditions. The FM contribution is enhanced as we increase the time of the thermal treatment in vacuum. Those systematic experiments establish a direct relation between ferromagnetism and oxygen vacancies and open a path for develo** materials with tailored properties. △ Less

Submitted 9 April, 2012; originally announced April 2012.

Comments: 20 pages, 3 figures; Applied Physics Letters Vol. 100, Issue 17, APR2012

Journal ref: Appl. Phys. Lett. 100, 172405 (2012)

Failure of the Local to Global Principle in the Eigencurve

Authors: Alexander G. M. Paulin

Abstract: For a cuspidal automorphic representation of GL2/Q associated to a modular form, the local and global Langlands correspondences are compatible at all finite places of Q. On the p-adic Coleman-Mazur eigencurve this principle can fail (away from p) under one of two conditions: on a generically principal series component where monodromy vanishes; or on a generically special component where the rati… ▽ More For a cuspidal automorphic representation of GL2/Q associated to a modular form, the local and global Langlands correspondences are compatible at all finite places of Q. On the p-adic Coleman-Mazur eigencurve this principle can fail (away from p) under one of two conditions: on a generically principal series component where monodromy vanishes; or on a generically special component where the ratio of the Satake parameters degenerates. We prove, under mild restrictive hypotheses, that such points are the intersection of generically principal series and special components. This is a geometric analogue of Ribet's level raising and lowering theorems. △ Less

Submitted 12 January, 2010; originally announced January 2010.