Faites connaître cette offre !
Reference : UMR5205-CATHERINE-007
Workplace : VILLEURBANNE
Date of publication : Thursday, July 25, 2019
Type of Contract : FTC Scientist
Contract Period : 12 months
Expected date of employment : 16 September 2019
Proportion of work : Full time
Remuneration : between 2600 and 3700 euros gross monthly according to experience
Desired level of education : 5-year university degree
Experience required : Indifferent
Physically-based rendering allows to simulate light interactions in a 3d scene, to obtain an image which simulates reality. In practice, its main uses are in cinematrography for image synthesis, or in architecture to
predict the behavior of light in a building. This simulation seeks to find light paths between a sensor (a camera or the eye of the observer) and light sources, that interact with the scene. These paths can be hard
to find, for instance when a scene is lit by a very indirect light that bounces many times. In that case, we use so-called Metropolis Hastings algorithms, which, once a path is found, tries to randomly "mutate" this path (i.e., alter it) to obtain a new similar path. This progressively
allows to explore all light paths that contribute to the scene lighting, that is, all paths joining light sources to the sensor. These mutations can however produce invalid paths (i.e., mutating the path introduces an obstacle). The goal of this postdoc position is to explore mutation
strategies that take into account a similarity measure between light paths, and to study optimal sampling for rendering. In particular, given a similarity metric between paths, we will be seek to randomly explore the unit ball given this metric. For that, we will either explore the
path space as a Riemannian manifold for which the metric should be determined, or we will deform the path space to make it Euclidean with techniques such as Multi-Dimensional Scaling, so as to simplify processing.
The project context is the ANR CALiTrOp, locally supervised by Nicolas Bonneel, Victor Ostromoukhov and David Coeurjolly in the Origami team of the LIRIS Lab, in collaboration with other partners of this ANR project
such as IRIT and INRIA Grenoble.
The postdoc will develop mathematical tools for rendering, as well as develop c++ code to be integrated in an existing 3d rendering engine, such as Mitsuba, PBRT or LuxRender.
The candidate should ideally have good knowledges in mathematics (in particular geometry and/or differential geometry, statistical methods, and mathematical tools such as integration, stochastic methods, and linear algebra), as well as physics (geometric optics). Good knowledge of C++ is required. A PhD in computer science, mathematics or physics is required.
Research Unit (UMR 5205), the InfoRmatic Laboratory in Image and Information Systems (LIRIS), LIRIS is supported by CNRS, INSA Lyon, University Claude Bernard Lyon 1, Université Lumière Lyon 2 and the Ecole Centrale de Lyon.
It has 330 members (including 134 EC / C, 125 PhD, 25 postdoctoral and ATER, 19 technical and administrative staff and 25 contractors, associates, Emerit), and its main field of science is computer science and more generally the Sciences et Technologies de 'Information.
A significant part of the research carried out at LIRIS extends to the border of our discipline, serving important societal issues. Some of our research activities are at the interfaces of engineering, humanities and social sciences, life sciences and environmental sciences. All of LIRIS '6 competence centers play a balanced role in the promotion of research work.
We talk about it on Twitter!