Ph.D. Position: Learning Non-rigid Surface Matching

Context

The Ph.D. will take place within the Morpheo research team at Inria Grenoble Rhône-Alpes. The team deals with the capture and analysis of dynamic scenes from multi-camera studios, and operates its own 68 camera acquisition platform and cluster, http://kinovis.inrialpes.fr.

 

[10]

 

The PhD topic is on surface matching which is  the process of finding correspondences between shape surfaces. Of particular interest whithin the Morpheo context is the application of surface matching to 3D human reconstructions with different body poses and dynamic clothing as captured  along time in dynamic mesh sequences. The objective is to achieve 4D temporally coherent meshes across complex dynamic scenes, e.g., mesh connectivity does not vary from frame-to-frame. The PhD will investigate learning based strategies for that purpose.

 

PhD Objectives

The focus of this Ph.D. is on matching unstructured surfaces from human performance capture systems, improving on traditional approaches [1, 2, 3, 4, 8] using novel sophisticated learning-based techniques [7, 9].  Learning-based techniques appear with considerable interest in shape analyses and representation [5, 6, 7]. Although most recent methods rely on statistical body models, which can lack on realistic surface deformations, in particular on clothing deformations. The direction of this research is towards enabling deep learning methods to efficiently represent non-rigid features, such as loose clothing and hair motion, consequently, facilitating matching of complex dynamic performance capture content.

The purpose of this Ph.D. is therefore to investigate innovative solutions on mesh matching that allow representation of non-rigid dynamic mesh sequences leverage by recent deep learning strategies that can take benefit of existing body shape datasets to learn non-rigid mesh correspondences.

 

Candidate profile

  • Master’s in computer science.
  • Background in 3D Vision and Machine Learning, knowledge in Computer Animation.
  • Solid programming skills, e.g. C++ and/or python.
  • Solid mathematical knowledge in linear algebra and statistics.
  • Good English level.

 

Position

The Ph.D. will start in October 2020 and its duration will be 3 years. The Ph.D. will be co-supervised by Joao Regateiro and Edmond Boyer at INRIA Grenoble.

 

Inria Grenoble
Inria is a leading French research centre in computer science, with an international culture – the English language being widely adopted. The Grenoble centre is located at the heart of the French Alps, a very dynamic region for new technologies offering a large range of recreational activities.

 

Application

Please apply on the INRIA website. Informal inquires can be addressed to joao-pedro.cova-regateiro@inria.fr and edmond.boyer@inria.fr .

 

References

[1] Chris Budd, Peng Huang, Martin Klaudiny, and Adrian Hilton. Global non-rigid alignment of surface sequences. International Journal of Computer Vision, 2013.

[2] C. Cagniart, E. Boyer, and S. Ilic. Free-form mesh tracking: A patch-based approach. In 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010.

[3] J. Regateiro, M. Volino, and A. Hilton. Hybrid skeleton driven surface registration for temporally consistent volumetric video. In 2018 International Conference on 3D Vision, 2018.

[4] Alvaro Collet, Ming Chuang, Pat Sweeney, Don Gillett, Dennis Evseev, David Calabrese, Hugues Hoppe, Adam Kirk, and Steve Sullivan. High-quality streamable free-viewpoint video, 2015.

[5] V. F. Abrevaya, A. Boukhayma, S. Wuhrer, and E. Boyer. A decoupled 3d facial shape model by adversarial training. In 2019 IEEE/CVF International Conference on Computer Vision, 2019.

[6] T. Alldieck, M. Magnor, B. L. Bhatnagar, C. Theobalt, and G. Pons-Moll. Learning to reconstruct people in clothing from a single rgb camera. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019.

[7] Verma, Nitika & Boyer, Edmond & Verbeek, Jakob. FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis, 2018.

[8] Mustafa, A. and Kim, H. and Guillemaut, J-Y. and Hilton, A. Temporally coherent 4D reconstruction of complex dynamic scenes, 2016.

[9] A. Caliskan, A. Mustafa, E. Imre and A. Hilton, “Learning Dense Wide Baseline Stereo Matching for People,” 2019.

[10] Avinash Sharma, Radu Horaud, Jan Cech, Edmond Boyer. Topologically-Robust 3D Shape Matching Based on Diffusion Geometry and Seed Growing, 2011.

Comments are closed.