PhD Proposal
UNIVERSITE MOHAMMED VI POLYTECHNIQUE - UM6P
"Machine learning and stochastic optimization algorithms applied to image synthesis" Supervisors: Fabien Teytaud, Julien Dehos, S ebastien Verel Laboratoire d'Informatique Signal et Image de la C^ote d'Opale (LISIC) Univ. of Littoral Cote d'Opale (ULCO), France. Keywords Machine learning, stochastic optimization, multi-armed bandit, image synthesis, multimodal func- tions, large scale learning and optimization. Scienti c description Image synthesis (also called rendering) consists in computing an image from a virtual 3D scene. It is a widespread technique used in many elds such as cinema, advertising, industrial design and so on. Rendering is a light transport simulation problem, de ned by a Fredholm integral equation called the rendering equation [Kaj86]. The rendering equation is di cult to solve, due to the large dimension and the distribution of the solutions, therefore, many numerical methods have been proposed. The Path-Tracing algorithm [Kaj86] is a straightforward Monte-Carlo method which success- fully solves the rendering equation. Many sampling techniques have been proposed to improve Path- Tracing, however the algorithm can still be very ine cient for scenes with complex illumination conditions [Big+18]. The Metropolis Light Transport algorithm [VG97] is an adaptation of the...
Morocco : Marrakech-Safi
2019-02-11
Contact our recruitment agency in Morocco