Projets proposés en 2022/2023
Résumé de section
-
-
MocapLab
(entreprise de capture de mouvement)
Contacts: Boris Dauriac (boris.dauriac@mocaplab.com), Rémi Brun (remi.brun@mocaplab.com)- Reconstruction d'un modèle 3D d’œil à partir de mesure ellipticité de l'iris et de la pupille depuis une vidéo
- Classification de mouvements de langue des signes française
- Benchmark de solution de lips-sync text+audio
Kinetix
(Startup reconstruction d'avatar 3D animé à partir de vidéo sur téléphone mobile)Contact: baptiste.b@kinetix.tech- Construction d'espaces latents pour l'animations de personnages.
Etude de l'article DeepPhase: Periodic Autoencoders for Learning Motion Phase Manifolds [github].Mercenaries Engineering/Rumba
(entreprise R&D logiciel Animation et Rendu)
Contacts: Cyril Corvazier (corvazier@mercs-eng.com), Lise Favre (favre@mercenaries-engineering.com), Florian Canezin (canezin@mercs-eng.com)
- Example-based Keyframe character interpolation.
Etudes des articles suivants:Concevoir d'un système qui, partant d'un personnage mis en position à certaines images clés de l'animation et d'une banque d'animation d'apprentissage, interpole ces positions clés en une animation complète.Elements fournis: licences Rumba, personnage, banque d'animations et implémentation débutante de l'algorithme dans un plug-in.Thales
- équipe analyse biométrique (bâtiment à coté du LIX):Contact: sandra.cremer@thalesgroup.com, sarah.lannes@thalesgroup.com- A survey on synthetic image generation methods to create faces : from closed-formed solutions to GANs and diffusion models. Creating safe, fair and representative datasets of artificial faces is one a the major challenges of the future of face biometrics, understanding the state-of-the-art is the first step towards that future.
- Creating an estimator for age and gender from a face image. Many aspects of image enhancement and face recognition rely at some point on this information as a way to gain robustness or accuracy. This project is about creating a python tool to get estimations from any image of a face, several datasets are available with labels associated.
- Face anti-spoofing detection. Any authentication solution suffers from attempts of fraud. In the case of face recognition, these frauds can range from presenting a picture or a video of someone else’s face to high-tech 3D masks impersonating someone else. This project is about implementing a state-of-the-art paper that has proposed a new approach to evaluate its potential.
LTCI/Telecom
Contact: Amal Dev Parrakat amal.parakkat@telecom-paris.fr(projet débutant par l'analyse de l'article mentionné)- Continuous line arts for understanding the geometry of a shape (Geometry, Creativity, Optimization) [article]
- Field generation from 3D meshes (Geometry) [article]
- Constrained mesh dilation (Geometry, Creativity) [article]
- Learning shape similarity for sketch-based image retrieval (Learning, Creativity, Sketching) [article]
- Image-guided 3D modelling (Creativity, Sketching, Modelling) [article]
- 3D reconstruction from sparse input (Geometry, Creativity) [article]
- Interactive colouring of animations (Applied Geometry, Creativity, maybe Learning) [article]
- Balancing 3D models (Applied physics/mechanics, optimization, geometry) [article]
- Stripe patterns on surfaces (Geometry) [article]
ModaLive
Startup - géneration de vêtements 3D complets à partir de scans.Contact: Thierry Godin (thierry@modalive.fr)- Lighting in augmented reality scenes
- Modeling of solid elements anchored in digitized clothing
- Roll up and tie virtual clothes
- Learning to measure the friction coefficient in clothes
LIX/Dassault Systèmes
- Interactive Modeling of 3D Shapes filled with Anisotropic Microstructure using Reaction Diffusion [link] (Damien Rohmer/LIX damien.rohmer@polytechnique.edu, David-Henri Garnier/Dassault Systèmes/LIX DavidHenri.GARNIER@3ds.com)
-