The field of 3D Gaussian Splatting (3DGS) is rapidly advancing, with a focus on improving scene reconstruction, novel view synthesis, and real-time rendering. Recent developments have addressed challenges such as sparse view reconstruction, multi-view consistency, and geometric accuracy. Notably, the integration of physics-driven light-interaction modeling and differentiable rendering-based reprojection strategies has enhanced the robustness of 3DGS methods. Additionally, the incorporation of diffusion models, material-guided relighting, and controllable stylized distillation has expanded the capabilities of 3DGS in various applications. Overall, the field is moving towards more efficient, accurate, and flexible 3DGS methods that can handle complex scenes and diverse lighting conditions. Noteworthy papers include: Optimization-Free Style Transfer for 3D Gaussian Splats, which proposes a reconstruction- and optimization-free approach to stylizing 3D Gaussian splats, and A 3DGS-Diffusion Self-Supervised Framework for Normal Estimation from a Single Image, which introduces a novel self-supervised framework for normal estimation via 3D Gaussian splatting guided diffusion.
Advancements in 3D Gaussian Splatting for Scene Reconstruction and Novel View Synthesis
Sources
Optimization-Free Style Transfer for 3D Gaussian Splats
A 3DGS-Diffusion Self-Supervised Framework for Normal Estimation from a Single Image
ExploreGS: Explorable 3D Scene Reconstruction with Virtual Camera Samplings and Diffusion Priors
UW-3DGS: Underwater 3D Reconstruction with Physics-Aware Gaussian Splatting
LightSwitch: Multi-view Relighting with Material-guided Diffusion
Evaluating Fisheye-Compatible 3D Gaussian Splatting Methods on Real Images Beyond 180 Degree Field of View
DIP-GS: Deep Image Prior For Gaussian Splatting Sparse View Recovery
GS4Buildings: Prior-Guided Gaussian Splatting for 3D Building Reconstruction
Novel View Synthesis with Gaussian Splatting: Impact on Photogrammetry Model Accuracy and Resolution
Make Your MoVe: Make Your 3D Contents by Adapting Multi-View Diffusion Models to External Editing
Multi-view Normal and Distance Guidance Gaussian Splatting for Surface Reconstruction
FantasyStyle: Controllable Stylized Distillation for 3D Gaussian Splatting
SAGOnline: Segment Any Gaussians Online
Communication Efficient Robotic Mixed Reality with Gaussian Splatting Cross-Layer Optimization
Gradient-Direction-Aware Density Control for 3D Gaussian Splatting
SkySplat: Generalizable 3D Gaussian Splatting from Multi-Temporal Sparse Satellite Images
DualPhys-GS: Dual Physically-Guided 3D Gaussian Splatting for Underwater Scene Reconstruction
Noise-adapted Neural Operator for Robust Non-Line-of-Sight Imaging
GSFixer: Improving 3D Gaussian Splatting with Reference-Guided Video Diffusion Priors
RayletDF: Raylet Distance Fields for Generalizable 3D Surface Reconstruction from Point Clouds or Gaussians
A Survey on 3D Gaussian Splatting Applications: Segmentation, Editing, and Generation
Multi-Sample Anti-Aliasing and Constrained Optimization for 3D Gaussian Splatting