Research map / Point clouds / Point cloud classification
FastPoint: Accelerating 3D Point Cloud Model Inference via Sample Point Distance Prediction
Donghyun Lee, Dawoon Jeong, Jae W. Lee, Hongil Yoon
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 |
|---|---|---|---|
| replace FPS with MDPS at inference (no retraining) | PointVector-L / PointMetaBase-L · S3DIS, ScanNet, SemanticKITTI | mIoU: 71.33 -> 71.37 (PV-L S3DIS); 50.91 -> 50.79 (PV-L SemanticKITTI) | No clear effect |
| replace FPS with random sampling | PointVector-L · SemanticKITTI | mIoU: 50.91 -> 38.46 | Hurt |
| replace FPS with grid sampling at inference only | PointVector-L · ScanNet | mIoU: 70.70 -> 69.61 | Hurt |
| add Redundancy-Free Neighbor Search on top of MDPS (ablation) | FastPoint · ScanNet, PointMetaBase-L | end-to-end speedup: 2.89x -> 3.38x | Improved |
| increase number of MDPS segments from 1 to 6 (ablation) | MDPS · ScanNet, PointVector-L | mIoU: 70.34 -> 70.63 | Improved |
| increase number of MDPS segments from 1 to 6 (ablation) | MDPS · ScanNet, PointVector-L | sampling time (ms): 18.62 -> 22.06 | Hurt |
| reduce initial FPS ratio p from 0.2 to 0.05 (ablation) | FastPoint · ScanNet, PointMetaBase-L | end-to-end speedup: 2.70x -> 3.82x | Improved |
| reduce initial FPS ratio p from 0.2 to 0.05 (ablation) | FastPoint · ScanNet, PointMetaBase-L | mIoU: 70.95 -> 70.83 | Hurt |
| use MLP instead of power function for curve estimator (ablation) | MDPS · SemanticKITTI | MAPE (%): 6.6118 -> 1.9318 | Improved |
| apply estimator trained on SemanticKITTI to ScanNet (ablation) | FastPoint · ScanNet, PointMetaBase-L | estimation error (%): 0.77 -> 2.24 | Hurt |
| apply data augmentation to validation data (ablation) | FastPoint · ScanNet, PointMetaBase-L | sampling quality (%): 99.35 -> 99.10 | No clear effect |
| combine FastPoint with QuickFPS for initial FPS iterations | FastPoint · S3DIS, ScanNet, SemanticKITTI | end-to-end speedup vs FPS: 2.55x -> 2.76x | Improved |
Closest papers by meaning
- Frugal 3D Point Cloud Model Training via Progressive Near Point Filtering and Fused Aggregation (ECCV 2024 · similarity 0.91)
- PLATYPUS: Progressive Local Surface Estimator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2025 · similarity 0.86)
- Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2024 · similarity 0.86)
- Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning (AAAI 2024 · similarity 0.85)
- HashPoint: Accelerated Point Searching and Sampling for Neural Rendering (CVPR 2024 · similarity 0.85)
- StraightPCF: Straight Point Cloud Filtering (CVPR 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.
