Research map / 3D vision / NeRF and radiance fields
Dynamic Scene Understanding Through Object-Centric Voxelization and Neural Rendering
Yanpeng Zhao, Yiwei Hao, Siyu Gao, Yunbo Wang, Xiaokang Yang
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 |
|---|---|---|---|
| add object-centric voxelization (ablation) | DynaVol-S · 3ObjRand | PSNR: 32.15 -> 32.58 | Improved |
| add object-centric voxelization (ablation) | DynaVol-S · 8ObjFall | PSNR: 30.76 -> 31.78 | Improved |
| add forward deformation network (ablation) | DynaVol-S · 8ObjFall | mIOU: 75.91 -> 85.41 | Improved |
| add CRF post-processing (ablation) | DynaVol-S · 3ObjRand | PSNR: 32.57 -> 32.58 | No clear effect |
| add multi-grids joint optimization stage (ablation) | DynaVol-S · 3ObjRand | PSNR: 30.48 -> 32.58 | Improved |
| add warmup stage before joint optimization (vs joint optimization from scratch) (ablation) | DynaVol-S · 3ObjRand | mIOU: 14.73 -> 90.60 | Improved |
| increase latent codes from N=5 to N=10 (ablation) | DynaVol-S · 8ObjFall | mIOU: 42.80 -> 85.41 | Improved |
| increase latent codes from N=10 to N=15 (ablation) | DynaVol-S · 8ObjFall | mIOU: 85.41 -> 85.46 | No clear effect |
| vary density threshold delta_den from 1 to 2 (ablation) | DynaVol-S · 8ObjFall | mIOU: 85.41 -> 88.41 | Mixed |
| add per-point RGB loss (alpha_p=0.01 vs 0) (ablation) | DynaVol-S · 3ObjRand | PSNR: 30.00 -> 32.58 | Improved |
| raise per-point RGB loss weight from 0.01 to 0.1 (ablation) | DynaVol-S · 3ObjRand | mIOU: 90.60 -> 31.61 | Hurt |
| add semantic features (DynaVol to DynaVol-S) | DynaVol-S · Real-world scene decomposition (average) | mIOU: 41.71 -> 88.95 | 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.
- DynaVol — DynaVol: Unsupervised Learning for Dynamic Scenes through Object-Centric Voxelization (ICLR 2024) · 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
- DynaVol: Unsupervised Learning for Dynamic Scenes through Object-Centric Voxelization (ICLR 2024 · similarity 0.96)
- Object-Centric Representation Learning with Generative Spatial-Temporal Factorization (NeurIPS 2021 · similarity 0.89)
- DORSal: Diffusion for Object-centric Representations of Scenes $\textit{et al.}$ (ICLR 2024 · similarity 0.87)
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.87)
- VoxelSplat: Dynamic Gaussian Splatting as an Effective Loss for Occupancy and Flow Prediction (CVPR 2025 · similarity 0.87)
- Gear-NeRF: Free-Viewpoint Rendering and Tracking with Motion-aware Spatio-Temporal Sampling (CVPR 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.
