Research map / 3D vision / Face and shape
SobolevFusion: 3D Reconstruction of Scenes Undergoing Free Non-Rigid Motion
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Sobolev gradient flow instead of the L2 gradient for TSDF warping | SobolevFusion · Mechanical toy sequences (KillingFusion) | geometric error: 3.9 mm -> 3.7 mm; 3.5 mm -> 3.1 mm | Helped |
| Use Sobolev gradient flow for TSDF warping | SobolevFusion · Kinect sequences with interacting subjects | geometric detail | Helped |
| Use Sobolev gradient flow for TSDF warping | SobolevFusion · Warped Umbrella frames | geometric detail | Helped |
| Use Sobolev gradient flow for TSDF warping | SobolevFusion · Alex sequence with reduced frame rate | recoverable frame speedup: 10-frame speedup; 15-frame speedup | Helped |
| Estimate correspondences using Laplacian eigenfunction signatures | SobolevFusion voxel matching · Vicon yt sequence | mean ℓ1 error: 7.7 cm vs 4.4 cm; 7.7 cm vs 3.7 cm | Hurt |
| Reduce Sobolev neighbourhood size to s = 3 (ablation) | SobolevFusion · Andrew-Chair sequence | reconstruction quality: s = 3 | Hurt |
| Remove motion regularization (ablation) | SobolevFusion · Andrew-Chair sequence | geometry consistency: wreg = 0 | Hurt |
| Increase the Sobolev filter parameter λ (ablation) | SobolevFusion · SobolevFusion convergence | iterations to converge: 3-8% | Helped |
| Use a Sobolev neighbourhood size s ≥ 7 rather than s = 3 (ablation) | SobolevFusion · SobolevFusion parameter sweep | reconstruction quality: s = 3; s ≥ 7 | Helped |
Closest papers by meaning
- KillingFusion: Non-Rigid 3D Reconstruction Without Correspondences (CVPR 2017 · similarity 0.90)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.85)
- ProbeSDF: Light Field Probes For Neural Surface Reconstruction (CVPR 2025 · similarity 0.85)
- DySurface: Consistent 4D Surface Reconstruction via Bridging Explicit Gaussians and Implicit Functions (NeurIPS 2026 · similarity 0.85)
- GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction (NeurIPS 2025 · similarity 0.85)
- Flow Matching on General Geometries (ICLR 2024 · similarity 0.85)
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.