Vous êtes ici

A topology-driven approach to retargeting of filmic style in 3D environments

Equipe et encadrants
Département / Equipe: 
Site Web Equipe: 
https://team.inria.fr/mimetic/
Directeur de thèse
Multon Franck
Co-directeur(s), co-encadrant(s)
Christie Marc
Lamarche Fabrice
Contact(s)
NomAdresse e-mailTéléphone
Christie Marc
marc.christie@irisa.fr
0650012922
Sujet de thèse
Descriptif

Context: 

Recent computational approaches have demonstrated how some elements of filmic style such as framing [1,3], camera motions [4], editing [1,6], or lighting may be formalized and transferred from existing data (real movies) to virtual 3D environments following the general objective of improving synthetic animated media creation for film industry, interactive stories and games.

However all current approaches only focus on the formalisation of one dimension of film style separately of others, like framing, editing or lighting. However caracterization of film style, in its general acception, can only be considered as a combination of its dimensions [7] typically including staging (how actors are positionned in the scene) – a dimension insufficiently addressed in the community.

Objectives:

The objective of this PhD is to first propose a formalization of filmic style which gathers staging, framing, lighting and their temporal evolution so as to represent information from real film footage in supervised ways. The challenge lies in the ability to express such complex dimensions in a generalized – and expressive – model which should ensure computational tractability. The PhD will design a film annotation language to analyse movies from existing approaches [8] and rely on the proposal of generalized models which encode the relations between cameras, lights and characters, informally described in the film literature.

The second objective is to then propose means to retarget extracted dimensions to adpat framing, lighting and staging to the constraints of a given 3D environment. The challenge consists in providing both spatial and temporal warping techniques to adapt positions and motions of cameras, lights and characters. Such techniques need to build on (i) a relevant topological analysis of the environment [2] so as to identify where are relevant locations in the 3D environment, and (ii) a hybrid action-and-geometric planning techniques to orchestrate the motions of characters (and hence of lights and cameras).

The proposed model should furthermore be designed in a way that these features of cinematographic style can be interactively controlled, which then opens the possibility not so much of replicating style, but creatively altering, blending and manipulating styles which after all is the essence of cinematography.

Bibliographie

[1] C Lino, M Christie, F Lamarche, G Schofield, P Olivier, « A real-time cinematography system for interactive 3d environments », Proceedings of the 2010 ACM SIGGRAPH/Eurographics Symposium on Computer Animation, 2010.

[2] F Lamarche, « Topoplan: a topological path planner for real time human navigation under floor and ceiling constraints », Computer Graphics Forum 28 (2), 649-658, 2010.

[3] C Sanokho, M Christie, « On-Screen Visual Balance Inspired by Real Movies », WICED Workshop at the Twenty-Eighth AAAI Conference on Artificial Intelligence, 2014.

[4] C Sanokho, C Desoche, B Merabti, TY Li, M Christie, « Camera motion graphs », Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation, 2014.

[5] Q Galvane, R Ronfard, C Lino, M Christie, « Continuity editing for 3D animation », AAAI Conference on Artificial Intelligence, 2015.

[6] B Merabti, M Christie, K Bouatouch, « A Virtual Director Using Hidden Markov Models », Computer Graphics Forum, 2015.

[7] D. Bordwell, « On the History of Film Style », Harvard University Press, 1993 

[8] HY Wu, M Christie, « Analysing Cinematography with Embedded Constrained Patterns » WICED-Eurographics Workshop on Intelligent Cinematography and Editing, 2016

Début des travaux: 
Octobre 2016
Mots clés: 
computational cinematography, film style, topology, retargeting
Lieu: 
IRISA - Campus universitaire de Beaulieu, Rennes