Research map / Point clouds / Shape and surface completion
SFD-Net: Sharp Feature Detection Network Based on Local Geometric Features
Inyoung Oh, Kwanghee Ko
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 LGD descriptor to coordinates (ablation) | SFD-Net · ABC (density variation only) | F1: 58.55 -> 61.33 | Improved |
| use nested scales k in {20,40,80} instead of {10,20,40} (ablation) | SFD-Net · ABC (density variation only) | F1: 61.16 -> 61.33 | Improved |
| concatenate per-scale descriptors instead of averaging (ablation) | SFD-Net · ABC (density variation only) | F1: 60.05 -> 61.33 | Improved |
| add dual SA + FP Transformer blocks (ablation) | SFD-Net · ABC (density variation only) | F1: 61.01 -> 61.33 | Improved |
| add Transformer at both SA and FP instead of FP only (ablation) | SFD-Net · ABC (density variation only) | F1: 60.02 -> 61.33 | Improved |
| add Transformer at both SA and FP instead of SA only (ablation) | SFD-Net · ABC (density variation only) | F1: 60.18 -> 61.33 | Improved |
| use composite Focal Tversky-Dice loss instead of Dice only (ablation) | SFD-Net · ABC (density variation only) | F1: 59.16 -> 61.33 | Improved |
| use composite loss instead of Focal Tversky only (ablation) | SFD-Net · ABC (density variation only) | F1: 59.93 -> 61.33 | Improved |
| use composite loss instead of Cross-Entropy (ablation) | SFD-Net · ABC (density variation only) | FPR: 40.85 -> 36.93 | Improved |
| prepend LGD to PointNet++ | PointNet++ · ABC (density variation only) | F1: 50.89 -> 57.89 | Improved |
| prepend LGD to RepSurf | RepSurf · ABC (density variation only) | F1: 55.23 -> 59.28 | Improved |
| prepend LGD to EDWG | EDWG · ABC (density variation only) | FPR: 46.06 -> 36.98 | Improved |
Closest papers by meaning
- Frustum PointNets for 3D Object Detection From RGB-D Data (CVPR 2018 · similarity 0.86)
- HEDNet: A Hierarchical Encoder-Decoder Network for 3D Object Detection in Point Clouds (NeurIPS 2023 · similarity 0.86)
- 3D Object Detection Using Scale Invariant and Feature Reweighting Networks (AAAI 2019 · similarity 0.86)
- Uni3DETR: Unified 3D Detection Transformer (NeurIPS 2023 · similarity 0.86)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.85)
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition (CVPR 2024 · similarity 0.84)
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.
