Research map / Point clouds / Point cloud classification
GeoMAE: Masked Geometric Target Prediction for Self-Supervised Point Cloud Pre-Training
Read the paperCheck 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 |
|---|---|---|---|
| pre-train SST with GeoMAE geometric targets | SST anchor-based detector · nuScenes validation | mAP: 50.39 -> 53.77 | Helped |
| pre-train SST with GeoMAE geometric targets | SST anchor-based detector · Waymo Open Dataset validation | L1 AP: 70.68 -> 73.71 | Helped |
| pre-train SST with GeoMAE | Cylinder3D semantic segmentation · nuScenes validation | mIoU: 76.5 -> 78.6 | Helped |
| pre-train SST with GeoMAE | Centerpoint tracking · nuScenes validation | AMOTA: 59.9 -> 61.6 | Helped |
| add centroid and occupancy prediction targets (ablation) | GeoMAE pre-training · Waymo Open Dataset | L1 AP: 70.68 -> 72.65 | Helped |
| add centroid and occupancy prediction targets (ablation) | GeoMAE pre-training · nuScenes | mAP: 50.39 -> 52.12 | Helped |
| add surface normal and curvature prediction targets (ablation) | GeoMAE pre-training · Waymo Open Dataset | L1 AP: 72.65 -> 73.71 | Helped |
| add surface normal and curvature prediction targets (ablation) | GeoMAE pre-training · nuScenes | mAP: 52.12 -> 53.31 | Helped |
| use separate decoders for different target groups instead of a shared decoder (ablation) | GeoMAE · Waymo Open Dataset | L1 AP: 72.45 -> 73.71 | Helped |
| use separate decoders for different target groups instead of a shared decoder (ablation) | GeoMAE · nuScenes | mAP: 52.04 -> 53.31 | Helped |
| set the masking ratio to 70% rather than 40%, 60%, or 80% (ablation) | GeoMAE pre-training · Waymo Open Dataset | L1 AP: 72.27; 73.28; 73.71; 71.94 | Mixed |
| extend pre-training from 24 to 72 epochs (ablation) | GeoMAE · Waymo Open Dataset | L1 AP: 72.67 -> 73.71 | Helped |
Closest papers by meaning
- 3D Feature Prediction for Masked-AutoEncoder-Based Point Cloud Pretraining (ICLR 2024 · similarity 0.91)
- RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learning (AAAI 2025 · similarity 0.90)
- PCP-MAE: Learning to Predict Centers for Point Masked Autoencoders (NeurIPS 2024 · similarity 0.88)
- Multi-Scale Neighborhood Occupancy Masked Autoencoder for Self-Supervised Learning in LiDAR Point Clouds (CVPR 2025 · similarity 0.88)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.88)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.88)
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.