Research map / 3D vision / 3D generation
Masked Space-Time Hash Encoding for Efficient Dynamic Scene Reconstruction
Feng Wang, Zilong Chen, Guokang Wang, Yafei Song, Huaping Liu
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 |
|---|---|---|---|
| replace the pure 4D hash encoding with the masked space-time hash encoding (ablation) | MSTH · dynamic scene reconstruction | PSNR: 3 PSNR drop with only 4D hash | Improved |
| replace simple addition of 3D and 4D hash tables with the uncertainty-guided mask (ablation) | MSTH · dynamic scene reconstruction | LPIPS: 20% | Improved |
| add uncertainty-based guidance to mask learning (ablation) | MSTH · mask quality / motion details | qualitative mask separation | Improved |
| use a simple addition of 3D and 4D tables instead of pure 4D hash (ablation) | hash encoding variant · dynamic scene reconstruction | reconstruction quality | Improved |
| MSTH versus prior methods on Plenoptic Video | MSTH · Plenoptic Video | PSNR: 0.7 to 1.4 PSNR gains | Improved |
| MSTH versus DyNeRF training time | MSTH · Plenoptic Video | training time: 1344 h -> 20 min | Improved |
| MSTH versus HyperReel on Google Immersive | MSTH · Google Immersive | PSNR: 28.8 -> 29.6 | Improved |
| MSTH versus TiNeuVox-B on D-NeRF | MSTH · D-NeRF | PSNR: 32.67 -> 31.34 | Hurt |
| MSTH versus TiNeuVox-B on D-NeRF | MSTH · D-NeRF | LPIPS: 0.04 -> 0.02 | Improved |
| MSTH versus K-Planes on Plenoptic Video | MSTH · Plenoptic Video | PSNR: 31.63 -> 32.37 | Improved |
| MSTH versus MixVoxels on storage | MSTH · Plenoptic Video | storage: 500MB -> 135MB | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HexPlane† — HexPlane: A Fast Representation for Dynamic Scenes (CVPR 2023) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DASH: 4D Hash Encoding with Self-Supervised Decomposition for Real-Time Dynamic Scene Rendering (ICCV 2025 · similarity 0.89)
- Grid4D: 4D Decomposed Hash Encoding for High-Fidelity Dynamic Gaussian Splatting (NeurIPS 2024 · similarity 0.88)
- Spatiotemporal-Aware Neural Fields for Dynamic CT Reconstruction (AAAI 2025 · similarity 0.84)
- Swift4D: Adaptive divide-and-conquer Gaussian Splatting for compact and efficient reconstruction of dynamic scene (ICLR 2025 · similarity 0.84)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.84)
- Hash3D: Training-free Acceleration for 3D Generation (CVPR 2025 · similarity 0.84)
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.
