Thesis proposal: Remote 3D visualisation from monoscopic views using Gaussian splats

Open Master's thesis · RISE, Kista · Start September 2026 · Contact ian.marsh@ri.se
Thesis: Remote 3D visualisation from monoscopic views using Gaussian splats

New to Gaussian splats? Read the one-page explainer, which has a truck scene trained on a laptop that you can orbit in your browser.

Thesis background

Remote-controlled moving platform (RCMP) setup: a truck with an ultrasonic sensor, a camera and a GPU sends video over a Wi-Fi access point to a receiver with a truck controller and a conventional 2D display; glass-to-glass latency 150 ms.
Figure 1: Remote-controlled moving platform (RCMP) setup.

We are offering a Master's thesis in digital twins, extended reality, robotics and autonomous systems, covering computer vision, 3D graphics and AI-driven scene representation. It focuses on the emerging field of 3D Gaussian splatting, which has gained attention as a powerful alternative to Neural Radiance Fields for high-quality, real-time rendering of complex 3D environments.

Gaussian splatting represents a scene as a collection of anisotropic Gaussian primitives. This gives photorealistic rendering with much better efficiency and interactivity. Your work will cover algorithm development, implementation, experiments and scene evaluation, using GPU-based frameworks.

Thesis focus

  • Real-time 3D scene reconstruction
  • Dynamic Gaussian splatting for moving objects and environments
  • Compression and optimisation of Gaussian representations

Work tasks

  • A state-of-the-art survey, including the code bases in this area
  • Practical work in Python/C++, CUDA and deep-learning frameworks
  • A demonstration on a moving platform: a 3D rendering of what a camera mounted on the platform sees, shown on a 3D display

Desired background

We are looking for very motivated students with interests in:

  • Computer vision and graphics
  • Machine learning
  • Experience with PyTorch, OpenGL, CUDA or 3D reconstruction is very desirable

Your background

You study at a Swedish university and have a solid technical background, including machine learning. You program in at least two languages: Python plus one of Java, Rust or C++.

Other information

  • You get an office at RISE in Kista and are expected there 2–3 days a week.
  • Pay is 1,333 SEK per ECTS credit, tax deductible, paid after a satisfactory oral and written thesis defence.
  • Start: September 2026.

Contact

Ian Marsh, PhD, Senior Researcher · ian.marsh@ri.se · Tel. +46 70 772 1536
Location: Kista
Union representatives: Sveriges Ingenjörer

Full proposal (Google Doc)

← All supervised theses