Research map / 3D vision / Face and shape
FastSESR: Fast Scene-level Explicit Surface Reconstruction
Jueqi Liu, Xuechao Zou, Congyan Lang
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 learnable OffsetFormer instead of SGD test-time optimization (T=10) (ablation) | FastSESR · CARLA | CD1 (x10^2): 6.715 -> 0.114 | Improved |
| use learnable OffsetFormer instead of SGD test-time optimization (T=10) (ablation) | FastSESR · CARLA | time (s): 269.94 -> 32.36 | Improved |
| use learnable OffsetFormer instead of SGD test-time optimization (T=2) (ablation) | FastSESR · CARLA | NR (degrees): 41.959 -> 4.929 | Improved |
| increase TTO steps from 2 to 10 (ablation) | FastSESR · CARLA | CD1 (x10^2): 6.745 -> 6.715 | No clear effect |
| use edge-factorized parameterization instead of dense pairwise prediction (ablation) | FastSESR · TCN parameter count | Params (M): 6.08 -> 0.35 | Improved |
| use edge-factorized parameterization instead of dense pairwise prediction (ablation) | FastSESR · ABC | NR (degrees): 1.305 -> 1.311 | No clear effect |
| replace test-time optimization (OffsetOPT) with feed-forward FastSESR | OffsetOPT · ScanNet | time per scene (s): 599.47 -> 24.43 | Improved |
| replace test-time optimization (OffsetOPT) with feed-forward FastSESR | OffsetOPT · Matterport3D | time per scene (s): 854.31 -> 39.04 | Improved |
| replace test-time optimization (OffsetOPT) with feed-forward FastSESR | OffsetOPT · CARLA | time per scene (s): 842.16 -> 32.36 | Improved |
| FastSESR vs OffsetOPT on geometric accuracy | OffsetOPT · ScanNet | CD1 (x10^2): 0.147 -> 0.154 | Hurt |
| FastSESR vs OffsetOPT on surface quality | OffsetOPT · CARLA | NR (degrees): 5.530 -> 4.929 | Improved |
Closest papers by meaning
- A Survey and Benchmark of Automatic Surface Reconstruction From Point Clouds (TPAMI 2024 · similarity 0.88)
- OffsetOPT: Explicit Surface Reconstruction without Normals (CVPR 2025 · similarity 0.87)
- VKSR: Scalable Kernel Surface Reconstruction Using Vecchia's Approximation (ECCV 2026 · similarity 0.87)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.86)
- High-Fidelity Lightweight Mesh Reconstruction from Point Clouds (CVPR 2025 · similarity 0.86)
- GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction (NeurIPS 2025 · 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.
