Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

4D RECONSTRUCTION IN SPARSE CAMERA SETTINGS FOR VR

Bati, Ezgi Aysel LU and Chen, Yiran LU (2026) MAMM15 20261
Department of Design Sciences
Ergonomics and Aerosol Technology
Abstract
This thesis explores 4D reconstruction in sparse fixed-camera settings for virtual reality visualization. Gaussian Splatting provides an efficient representation for static and dynamic scenes, but most existing pipelines rely on dense multi-view capture or sufficient camera motion. This project focuses on a constrained setting with only four fixed cameras observing a dynamic human object, where limited view coverage makes stable geometry and novel-view rendering difficult. We first evaluate vanilla 3DGS and 4DGS pipelines on public datasets and self-recorded data to identify their limitations under sparse fixed views. The baseline experiments show that the original pipelines struggle to obtain reliable camera poses and stable geometry.... (More)
This thesis explores 4D reconstruction in sparse fixed-camera settings for virtual reality visualization. Gaussian Splatting provides an efficient representation for static and dynamic scenes, but most existing pipelines rely on dense multi-view capture or sufficient camera motion. This project focuses on a constrained setting with only four fixed cameras observing a dynamic human object, where limited view coverage makes stable geometry and novel-view rendering difficult. We first evaluate vanilla 3DGS and 4DGS pipelines on public datasets and self-recorded data to identify their limitations under sparse fixed views. The baseline experiments show that the original pipelines struggle to obtain reliable camera poses and stable geometry. Geometry-based improvements are then explored, including depth-prior supervision and cross-view geometric consistency. These constraints provide additional supervision, but they do not recover unobserved regions or produce a stable full 360° reconstruction. Based on these observations, the final pipeline adopts a generative completion approach. The four cameras are calibrated, foreground masks are generated to isolate the target human object, and Diffuman4D is used to synthesize novel-view human videos. The generated multi-view data is adapted to vanilla 4DGS and trained with modified data loading to handle the larger dataset. The dynamic Gaussian representation is then exported as per-frame splat files and visualized in a Unity-based VR environment. The results show that generative completion is more suitable than geometry-only stabilization for this setting. The final reconstruction enables recognizable 360° dynamic observation, but remains limited by temporal inconsistencies in generated images and residual foreground noise. (Less)
Please use this url to cite or link to this publication:
author
Bati, Ezgi Aysel LU and Chen, Yiran LU
supervisor
organization
course
MAMM15 20261
year
type
H2 - Master's Degree (Two Years)
subject
keywords
4DGS, Gaussian Splatting, Novel View Synthesis, Virtual Reality
language
English
id
9230317
date added to LUP
2026-06-04 10:21:33
date last changed
2026-06-04 10:21:33
@misc{9230317,
  abstract     = {{This thesis explores 4D reconstruction in sparse fixed-camera settings for virtual reality visualization. Gaussian Splatting provides an efficient representation for static and dynamic scenes, but most existing pipelines rely on dense multi-view capture or sufficient camera motion. This project focuses on a constrained setting with only four fixed cameras observing a dynamic human object, where limited view coverage makes stable geometry and novel-view rendering difficult. We first evaluate vanilla 3DGS and 4DGS pipelines on public datasets and self-recorded data to identify their limitations under sparse fixed views. The baseline experiments show that the original pipelines struggle to obtain reliable camera poses and stable geometry. Geometry-based improvements are then explored, including depth-prior supervision and cross-view geometric consistency. These constraints provide additional supervision, but they do not recover unobserved regions or produce a stable full 360° reconstruction. Based on these observations, the final pipeline adopts a generative completion approach. The four cameras are calibrated, foreground masks are generated to isolate the target human object, and Diffuman4D is used to synthesize novel-view human videos. The generated multi-view data is adapted to vanilla 4DGS and trained with modified data loading to handle the larger dataset. The dynamic Gaussian representation is then exported as per-frame splat files and visualized in a Unity-based VR environment. The results show that generative completion is more suitable than geometry-only stabilization for this setting. The final reconstruction enables recognizable 360° dynamic observation, but remains limited by temporal inconsistencies in generated images and residual foreground noise.}},
  author       = {{Bati, Ezgi Aysel and Chen, Yiran}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{4D RECONSTRUCTION IN SPARSE CAMERA SETTINGS FOR VR}},
  year         = {{2026}},
}