PRIM Topics in Computer Graphics
IGR Filière, 2022-23

Students carry out a research project of interesting research articles. The project will be supervised by one of our experts in the computer graphics field. In case of PRIM, it is also possible to carry out a project in a company, co-supervised by a teacher from our school, with a topic that has been validated by the teacher. (No PRIM project will be possible without the validation of the teacher).

The project will be evaluated by submitted work including a report and final defense. The report must be submitted by the student before the defense. The members of the Computer Graphics group may attend the defense and participate in the Q/A session. The evaluation criteria are not negotiable even in the topic from a company and even if there is confidential contents. If any constraints apply to the topic, the student must notify the host company and teacher before the project is accepted.

Topics

IMPORTANT: Please be aware of that the topics listed below will be flexibly adapted according to your background knowledge and interests. Moreover, in addition to the topics listed below, you can still discuss other topics with your potential supervisor. Please feel free to contact supervisors (listed at each topic below) if you have any interesting directions you want to investigate in the computer graphics area.

This project aims to implement a stable fluid solver using machine learning library, TensorFlow, such that the solver is differentiable and can be integrated in training process. This topic requires good understanding of partial differential equations, in particular, the Navier-Stokes equations that model a variety of fluid flows. Moreover, it requires nice programming skills (particularly in Python). You should be able to make good use of TensorFlow.
Dreamcities 1 — Supervisors: Jérôme DUVAL and Jano HA, (Contact: Daniela Cirigliano Peschard)
Le projet est supporté par un moteur temps réel de type Unity. L’ensemble des quartiers de villes est reproduit par un cumul de 3D temps réel, de photogrammétries et de vidéos. Le tout retransmis par un SDK VR, dans le casque VR. La di usion étant assurée en CloudXR via notre partenaire. Il s’agit d’étudier et d’optimiser les transferts de ux vidéo, depuis le moteur temps réel vers la compression stéréoscopique en VR. Étudier "l’état de l’art », et comparer les choix techniques possibles a n d’obtenir pour l’expérience utilisateur un ux supérieur à 80 FPS en toute circonstance.
Dreamcities 2 — Supervisors: Jérôme DUVAL and Jano HA, (Contact: Daniela Cirigliano Peschard)
Il s’agit d’assurer et garantir la sécurité des « builds » que nous générons. Et d’éviter qu’ils puissent être « ouverts », copiés, reproduits, et utilisés dans l’autorisation du serveur distant. Il s’agira pour vous de soit optimiser notre solution existante ( fragile selon nous ) soit d’en proposer une autre, et d’en e ectuer le code qui convient. ( Majoritairement en C# ).
Please see the pdf file.
TBD
...