Stage 4D Gaussian splatting for autonomous driving-Saclay-

Mathématiques, information  scientifique, logiciel Stage 4D Gaussian splatting for autonomous driving-Saclay- H/F 4D Gaussian splatting for autonomous driving As an intern at the CEA, you will have the opportunity to work in a world-renowned research environment. Our teams consist of passionate and dedicated experts, providing an environment conducive to learning and collaboration. You will have access to state-of-the-art equipment and top-tier research resources to carry out your assignments. The work performed may potentially lead to a scientific publication. Context Please read the attached file for proper formatting and illustration.     Autonomous driving requires long term understanding of the 3D driving environment in order to correctly perceive both its coarse general structure (HD map, road topology) and its fine constituting elements (vehicles, pedestriand, traffic signs ...) .     In this internship, we propose to generate gaussians from multimodal inputs (camera, LIDAR, RADAR) . These gaussians will then be used as an input that contains both global and fine-grained context necessary for 3D perception in the scene. The candidate will then adapt 3D perception algorithms to this input modality in order to achieve SOTA performance on perception tasks such as 3D detection and HD map estimation.What do we expect from you?   To achieve these objectives, the intern will be expected to: - Review the state of the art on gaussian splatting reconstruction and 3D perception - Design, develop and evaluate a novel deep learning pipeline for perception in autonomous driving scenes - Contribute to research reports and potential publications References:[1] LU, Yiren, YE, Xin, YAMAN, Burhaneddin, et al. Reconstruction Matters: Learning Geometry-Aligned BEV Representation through 3D Gaussian Splatting. arXiv preprint arXiv:2603.19193, 2026. [2] CHABOT, Florian, GRANGER, Nicolas, et LAPOUGE, Guillaume. Gaussianbev: 3d gaussian representation meets perception models for bev segmentation. In : 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). IEEE, 2025. p. 2250-2259. [3] KERBL, Bernhard, KOPANAS, Georgios, LEIMKÜHLER, Thomas, et al. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 2023, vol. 42, no 4, p. 139:1-139:14. #Cea List PYTHON, Pytorch, CUDA, proprietary software Profile :Students in their 4th or 5th year of studies (M1, M2 or gap year) Computer vision skills Machine learning skills (deep learning, perception models, generative AI…) Python proficiency in a deep learning framework (especially PyTorch or TensorFlow) Strong interest for 3D