Research map / 3D vision / Gaussian splatting
A Compact Dynamic 3D Gaussian Representation for Real-Time Dynamic View Synthesis
Kai Katsumata, Duc Minh Vo, Hideki Nakayama
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| model Gaussian position and rotation as low-parameter functions of time instead of per-timestep parameters (D-3DGS) | 3D Gaussian Splatting · D-NeRF | PSNR: 17.22 -> 32.19 | Improved |
| model Gaussian position and rotation as functions of time instead of per-timestep parameters (D-3DGS) | 3D Gaussian Splatting · DyNeRF | memory: ~2.3GB -> ~338MB | Improved |
| replace NeRF-based dynamic methods with dynamic 3D Gaussians | K-Planes · D-NeRF | FPS: 0.54 -> 150 | Improved |
| replace NeRF methods with dynamic 3D Gaussians on multi-view data | K-Planes · DyNeRF | PSNR: 31.63 -> 30.46 | Hurt |
| use Fourier position approximation with L=2 instead of L=5 (ablation) | compact dynamic 3DGS · D-NeRF (mean) | PSNR: 28.10 -> 32.19 | Improved |
| increase Fourier harmonics L per scene (ablation) | compact dynamic 3DGS · D-NeRF (Jumping Jacks, T-Rex) | PSNR: 30.95 -> 32.93 (Jacks, L=2 -> L=4) | Mixed |
| replace Fourier position approximation with polynomials (ablation) | compact dynamic 3DGS · D-NeRF (mean) | PSNR: 32.19 -> 25.37 (linear) | Hurt |
| replace Fourier position approximation with 5-point spline (ablation) | compact dynamic 3DGS · D-NeRF (mean) | PSNR: 32.19 -> 32.10 | No clear effect |
| add time-varying scale with linear approximation (ablation) | compact dynamic 3DGS · D-NeRF (mean) | PSNR: 32.19 -> 31.94 | No clear effect |
| add optical flow supervision loss (ablation) | compact dynamic 3DGS · DyNeRF | visual quality | Improved |
Closest papers by meaning
- Spacetime Gaussian Feature Splatting for Real-Time Dynamic View Synthesis (CVPR 2024 · similarity 0.94)
- Swift4D: Adaptive divide-and-conquer Gaussian Splatting for compact and efficient reconstruction of dynamic scene (ICLR 2025 · similarity 0.92)
- Fully Explicit Dynamic Gaussian Splatting (NeurIPS 2024 · similarity 0.92)
- Motion Decoupled 3D Gaussian Splatting for Dynamic Object Representation (AAAI 2025 · similarity 0.92)
- Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction (CVPR 2024 · similarity 0.92)
- Efficient Gaussian Splatting for Monocular Dynamic Scene Rendering via Sparse Time-Variant Attribute Modeling (AAAI 2025 · similarity 0.92)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
