Research map / 3D vision / Depth and stereo
Taylor-Gaussians-Flow: Towards Non-uniform Motion for Novel View Synthesis from Monocular Video
Zaoming Yan, Qizhou Chen, Yaomin Huang, Pengcheng Lei, Chenhao Shi, Yi Xu, Haichuan Song, Faming Fang
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 |
|---|---|---|---|
| use Taylor-Gaussians head with acceleration instead of velocity-only head (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 22.00 -> 25.19 | Improved |
| add second-order (acceleration) Taylor term (ablation) | TGs-Flow · SNU-HARD 8 frames | PSNR: 23.28 -> 25.19 | Improved |
| add first-order term over zero-order (ablation) | TGs-Flow · SNU-HARD 8 frames | PSNR: 19.71 -> 23.28 | Improved |
| add third-order supervision (ablation) | TGs-Flow · SNU-HARD 8 frames | stability | Hurt |
| add acceleration loss to RGB and flow losses (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 23.70 -> 24.86 | Improved |
| add optical-flow loss to RGB loss (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 20.33 -> 23.70 | Improved |
| add isotropic regularization (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 24.86 -> 25.19 | Improved |
| vary lambda_reg from 0.00 to 0.20 (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 25.13 (0.00), 25.10 (0.10), 25.06 (0.20) | No clear effect |
| tune lambda_f and lambda_a (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 24.69 to 25.10 | No clear effect |
| train optical flow estimator on more datasets (ablation) | FF-TGsF · SNU-HARD 8 frames | PSNR: 23.61 -> 28.63 | Improved |
| add second-order term to flow supervision (ablation) | TGs-Flow · Sintel final test | EPE: 2.57 -> 2.04 | Improved |
| use FF-TGsF instead of NoPoSplat on RE10K novel view | FF-TGsF · RE10K novel view | PSNR: 28.65 -> 26.63 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- NeoVerse — NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos (CVPR 2026) · better on 9 of 9 shared results
- MoVieS — MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Second (CVPR 2026) · better on 3 of 4 shared results
- StreamSplat — StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams (ICLR 2026) · better on 1 of 1 shared result
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
- World from Motion: Generative Dynamic Gaussian Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.87)
- HAIF-GS: Hierarchical and Induced Flow-Guided Gaussian Splatting for Dynamic Scene (NeurIPS 2025 · similarity 0.86)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · similarity 0.86)
- DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos (NeurIPS 2025 · similarity 0.86)
- HiCoM: Hierarchical Coherent Motion for Dynamic Streamable Scenes with 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.86)
- Gaussian-Flow: 4D Reconstruction with Dynamic 3D Gaussian Particle (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.
