Research map / Point clouds / Point cloud classification
FPConv: Learning Local Flattening for Point Convolution
Check 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 |
|---|---|---|---|
| Fuse FPConv with KPConv at the final feature level (ablation) | S3DIS Area-5 segmentation · S3DIS Area-5 | mIoU: 62.8 -> 66.7 | Helped |
| Fuse FPConv with PointConv at the convolutional feature level (ablation) | ScanNet segmentation · ScanNet | mIoU: 63.0 -> 64.2 | Helped |
| Fuse FPConv with PointConv at the final feature level (ablation) | S3DIS Area-5 segmentation · S3DIS Area-5 | mIoU: 62.8 -> 64.4 | Helped |
| Fuse KPConv with PointConv at the final feature level (ablation) | S3DIS Area-5 segmentation · S3DIS Area-5 | mIoU: 65.4 -> 65.5 | No clear effect |
| Use sparse normalization instead of no normalization (ablation) | FPConv · S3DIS Area-5 | mIoU: 59.8 -> 62.8 | Helped |
| Use sparse normalization instead of dense normalization (ablation) | FPConv · S3DIS Area-5 | mIoU: 61.6 -> 62.8 | Helped |
| Use a 6x6 plane instead of a 5x5 plane (ablation) | FPConv with sparse normalization · S3DIS Area-5 | mIoU: 61.8 -> 62.8 | Helped |
| Halve the middle channel size in the FPConv–PointConv fusion block (ablation) | FPConv ⊗ PointConv · ScanNet | mIoU: 64.2 -> 65.1 | Helped |
| Fuse different convolution types rather than two PointConv branches (ablation) | ScanNet · Convolution-level fusion | mIoU: 60.7 -> 64.2 | Helped |
| Fuse different convolution types rather than two FPConv branches (ablation) | ScanNet · Convolution-level fusion | mIoU: 62.9 -> 64.2 | Helped |
| Use surface-style FPConv instead of volumetric-style methods (ablation) | Point cloud segmentation across regions of different curvature · S3DIS | accuracy | Mixed |
Closest papers by meaning
- PointConv: Deep Convolutional Networks on 3D Point Clouds (CVPR 2019 · similarity 0.89)
- KPConvX: Modernizing Kernel Point Convolution with Kernel Attention (CVPR 2024 · similarity 0.87)
- StraightPCF: Straight Point Cloud Filtering (CVPR 2024 · similarity 0.87)
- PLATYPUS: Progressive Local Surface Estimator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2025 · similarity 0.87)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.86)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · 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.