Research map / Point clouds / Point cloud classification
Fully-Convolutional Point Networks for Large-Scale Point Clouds
Dario Rethage, Johanna Wald, Jurgen Sturm, Nassir Navab, Federico Tombari
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 FCPN instead of voxel-based ScanNet network | semantic voxel labeling · ScanNet semantic voxel labeling | weighted accuracy: 73.0% -> 82.6% | Improved |
| use FCPN instead of voxel-based ScanNet network | semantic voxel labeling · ScanNet semantic voxel labeling | unweighted accuracy: 50.8% -> 54.2% | Improved |
| use FCPN instead of PointNet++ MSG+DP | semantic voxel labeling · ScanNet semantic voxel labeling | weighted accuracy: 84.5% -> 82.6% | Hurt |
| use FCPN instead of PointNet++ MSG+DP | semantic voxel labeling · ScanNet semantic voxel labeling | unweighted accuracy: 60.2% -> 54.2% | Hurt |
| use FCPN instead of ScanNet baseline for chairs | semantic voxel labeling · ScanNet semantic voxel labeling - Chair | class accuracy: 69.3% -> 81.6% | Improved |
| use FCPN instead of ScanNet baseline for desks | semantic voxel labeling · ScanNet semantic voxel labeling - Desk | class accuracy: 36.8% -> 58.5% | Improved |
| use FCPN instead of ScanNet baseline for doors | semantic voxel labeling · ScanNet semantic voxel labeling - Door | class accuracy: 48.9% -> 16.6% | Hurt |
| use FCPN instead of PointNet++ on part segmentation | part segmentation · ShapeNet part segmentation | mean IoU: 85.1 -> 84.0 | Hurt |
| use FCPN instead of PointNet on part segmentation | part segmentation · ShapeNet part segmentation | mean IoU: 83.7 -> 84.0 | Improved |
| use FCPN instead of PointNet++ for cars | part segmentation · ShapeNet part segmentation - car | IoU: 77.3 -> 88.3 | Improved |
| use FCPN instead of PointNet++ for tables | part segmentation · ShapeNet part segmentation - table | IoU: 82.6 -> 73.4 | Hurt |
| scale input from 15k points (16 m2) to 150k points (80 m2) | FCPN inference · Titan Xp inference | memory: 6481 MB -> 9033 MB | Mixed |
Closest papers by meaning
- PointCNN++: Performant Convolution on Native Points (CVPR 2026 · similarity 0.93)
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.92)
- A-CNN: Annularly Convolutional Neural Networks on Point Clouds (CVPR 2019 · similarity 0.92)
- PointConv: Deep Convolutional Networks on 3D Point Clouds (CVPR 2019 · similarity 0.91)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.91)
- Deep FusionNet for Point Cloud Semantic Segmentation (ECCV 2020 · similarity 0.90)
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.
