InternCompression Of 3D Gaussian Splat Representations Using Video-Based And 3D-Based Techniques H/F

Job description

Description

Summary In recent years, several volumetric scene representations have been introduced to enable New View Generation (NVG)-the process of creating novel viewpoints from a limited number of captured images. Among these, 3D Gaussian Splatting (3DGS) has emerged as a state-of-the-art technique, delivering highly realistic and efficient rendering of 3D scenes. This approach represents a major evolution in how volumetric data is modeled. However, the high visual fidelity of 3DGS comes with a significant data overhead, making compression a key challenge in its practical deployment. This internship will center on the design and implementation of software solutions for compressing 3DGS data using advanced, state-of-the-art techniques. Two primary strategies will be investigated: Video-based compression, which involves projecting 3DGS attributes onto a 2D image plane and applying conventional video codecs. 3D-based compression, which adapts existing 3D codecs to accommodate the unique properties of 3DGS representations. Working closely with the InterDigital research team, the candidate will: Gain a comprehensive understanding of the compression pipeline for 3D Gaussian Splat sequences, including their data structures and core algorithmic components. Investigate and compare compression approaches based on 2D video codecs and 3D geometry codecs. Develop and assess various 3DGS compression methods across multiple datasets. Explore hybrid strategies that combine video-based and 3D-based techniques, and analyze their effectiveness. This internship provides a hands-on opportunity to engage in cutting-edge work at the crossroads of software engineering, 3D graphics, and applied research. The selected candidate will join the 3D Native Codec team, part of the New Media Codecs Group within InterDigital's Video Lab, contributing to innovative solutions in immersive media compression. Responsibilities In collaboration with the InterDigital research team, the candidate will be expected to: Analyze existing technologies and identify core challenges related to 3D Gaussian Splat compression. Contribute to the design and development of software solutions addressing these challenges. Implement, test, and validate the proposed components and algorithms. Clearly document findings and actively participate in internal presentations and technical discussions. Keywords: 3D Gaussian splatting, compression, immersive Expected Outcomes: Software modules: C++ functions, python scripts Standardization contributions Patents

Date de début

24 oct., 2025

Profil

Qualifications Currently enrolled in a Master's program in Computer Engineering, Computer Science, Software Engineering, or a related discipline, with a strong interest in applied research. Solid programming skills in C++ and Python (required). Prior experience or familiarity with data compression, 3D graphics, geometry processing, or computer vision is a strong asset. Fluent in English, with excellent written and verbal communication skills. Able to work effectively within a multidisciplinary research team and contribute to technical documentation and presentations.

Répartition du temps de travail

Full time

Durée (Mois)

6

Formation

RJ/Qualif/Ingenieur_B5

Secteur

Ind_hightech_telecom