Research map / 3D vision / 3D generation
View Synthesis of Dynamic Scenes Based on Deep 3D Mask Volume
Kai‐En Lin, Guowei Yang, Lei Xiao, Feng Liu, Ravi Ramamoorthi
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 2D mask with 3D mask volume (ablation) | MPI blending pipeline · Real multi-view video dataset (cameras 4,5 input) | STRRED: 0.2892 -> 0.1683 | Improved |
| replace 2D mask with 3D mask volume (ablation) | MPI blending pipeline · Real multi-view video dataset (cameras 4,5 input) | PSNR: 25.50 -> 26.22 | Improved |
| add 3D mask volume with background MPI to per-frame MPI/LLFF | MPI/LLFF · Real multi-view video dataset (cameras 4,5 input) | STRRED: 0.2917 -> 0.1683 | Improved |
| replace 2D mask with 3D mask volume (ablation) | MPI blending pipeline · Synthetic dataset (cameras 4,5 input) | STRRED: 0.5428 -> 0.2812 | Improved |
| add 3D mask volume to MPI/LLFF | MPI/LLFF · Synthetic dataset (cameras 4,5 input) | STRRED: 0.2889 -> 0.2812 | No clear effect |
| add 3D mask volume to MPI/LLFF | MPI/LLFF · Synthetic dataset (cameras 4,5 input) | PSNR: 26.1167 -> 26.1348 | No clear effect |
| use 2D mask instead of no mask (ablation) | MPI/LLFF · Real multi-view video dataset (cameras 4,5 input) | STRRED: 0.2917 -> 0.2892 | No clear effect |
| add mask sparsity loss Ls (ablation) | Ours · Real multi-view video dataset (cameras 4,5 input) | STRRED: 0.1683 -> 0.1745 | Hurt |
| add sparsity loss Ls and mask supervision loss Lm (ablation) | Ours · Real multi-view video dataset (cameras 4,5 input) | STRRED: 0.1683 -> 0.1900 | Hurt |
| add mask sparsity loss Ls (ablation) | Ours · Real multi-view video dataset (cameras 4,5 input) | SSIM: 0.8390 -> 0.8393 | No clear effect |
| use a 3D mask instead of NeRF-based dynamic methods on two input views | NSFF/NeRFlow/Non-rigid NeRF · Real multi-view video dataset (cameras 4,5 input) | PSNR: 17.0368 -> 26.22 | Improved |
| extend to 4 input views | Ours · qualitative | visual artifacts | Improved |
Closest papers by meaning
- Geometry-guided Online 3D Video Synthesis with Multi-View Temporal Consistency (CVPR 2025 · similarity 0.88)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.88)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 2025 · similarity 0.88)
- SimVS: Simulating World Inconsistencies for Robust View Synthesis (CVPR 2025 · similarity 0.87)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.87)
- Pseudo-Generalized Dynamic View Synthesis from a Video (ICLR 2024 · similarity 0.87)
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.
