PhD CIFRE, PhD – Next-generation realistic avatar representation for videoconferencing

Investigate Gaussian-avatar representations for multi-user immersive telepresence by extending current face-centric coding to unified full-body and facial encoding. Explore geometry-aware 3D Gaussian Splatting for view-dependent rendering of realistic head models while maintaining compatibility with skeletal joints and blendshape weights. Develop semantic, standard-compatible control mechanisms that avoid per-frame neural inference to ensure deterministic animation for real-time communication. Study the integration of facial and body components, handling secondary elements like hair and clothing under varying network constraints. Evaluate robustness, compression, and perceptual quality of Gaussian avatars during bandwidth-limited transmission. Collaborate with Inria teams to advance videoconference and metaverse applications through industrial patents and peer-reviewed publications.