Research map / 3D vision / NeRF and radiance fields
Time of the Flight of the Gaussians: Optimizing Depth Indirectly in Dynamic Radiance Fields
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Adapts dynamic 3D Gaussian splatting to reconstruct scenes from raw continuous-wave time-of-flight camera quads, using occupancy and low-reflectivity heuristics to make indirectly optimized depth accurate, ~100x faster than neural ToF radiance fields.
Problem
unused
Approach
unused
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the proposed C-ToF Gaussian splatting method instead of the evaluated baselines | Dynamic monocular C-ToF reconstruction · F-TöRF synthetic dataset | depth error | Mixed |
| Use Gaussian splatting with the proposed optimization rather than neural volumetric reconstruction | Dynamic C-ToF radiance fields · 4D scene reconstruction | optimization time: 60 hours; 72 hours; 40–60 minutes | Helped |
| Use the combined occupancy and low-reflectivity biases | Proposed method · Test sequences | depth accuracy | Helped |
| Remove the low-reflectivity bias while retaining the occupancy bias (ablation) | Proposed method · Low-reflectivity scene regions | depth accuracy | Hurt |
| Reduce the reflectivity learning rate 10× to favor occupancy changes | Proposed method · C-ToF radiance field reconstruction | depth accuracy: 10× | Helped |
| Initialize Gaussian reflectivity to a lower value (0.01 to 0.1) rather than a high value (e.g., 0.5) | Proposed method · Low-reflectivity scene regions | depth accuracy: 0.01 to 0.1; 0.5 | Helped |
Closest papers by meaning
- Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction (CVPR 2024 · similarity 0.88)
- Snapshot Lidar: Fourier Embedding of Amplitude and Phase for Single-Image Depth Reconstruction (CVPR 2024 · similarity 0.86)
- Spacetime Gaussian Feature Splatting for Real-Time Dynamic View Synthesis (CVPR 2024 · similarity 0.86)
- SC-OmniGS: Self-Calibrating Omnidirectional Gaussian Splatting (ICLR 2025 · similarity 0.86)
- HyRF: Hybrid Radiance Fields for Memory-efficient and High-quality Novel View Synthesis (NeurIPS 2025 · similarity 0.86)
- 4D Gaussian Splatting in the Wild with Uncertainty-Aware Regularization (NeurIPS 2024 · similarity 0.86)
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.