Research map / Point clouds / Point cloud classification
PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies
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 |
|---|---|---|---|
| Apply improved training strategies without changing architecture | PointNet++ · ScanObjectNN PB_T50_RS | OA: 77.9 -> 86.1 | Helped |
| Apply improved training strategies without changing architecture | PointNet++ · S3DIS 6-Fold | mIoU: 54.5 -> 68.1 | Helped |
| Add point resampling (ablation) | PointNet++ · ScanObjectNN PB_T50_RS | OA: +2.5 | Helped |
| Remove jittering (ablation) | PointNet++ · ScanObjectNN PB_T50_RS | OA: +1.1 | Helped |
| Add height appending (ablation) | PointNet++ · ScanObjectNN PB_T50_RS | OA: +1.1 | Helped |
| Add label smoothing (ablation) | PointNet++ · ScanObjectNN PB_T50_RS | OA: +1.3 | Helped |
| Add color drop (ablation) | PointNet++ · S3DIS Area-5 | mIoU: +5.9 | Helped |
| Use random rotation (ablation) | PointNet++ · S3DIS Area-5 | mIoU: -0.3 | Hurt |
| Add random rotation (ablation) | PointNeXt-B · S3DIS Area-5 | mIoU: +1.5 | Helped |
| Normalize relative coordinates by the neighborhood radius (ablation) | PointNet++ · ScanObjectNN PB_T50_RS | OA: +0.3 | Helped |
| Remove the residual connection (ablation) | PointNeXt-XL · S3DIS Area-5 | mIoU: 70.5 -> 64.0 | Hurt |
| Remove separable MLPs (ablation) | PointNeXt-XL · S3DIS Area-5 | mIoU: 70.5 -> 66.6 | Hurt |
Closest papers by meaning
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.91)
- Revisiting Point Cloud Shape Classification with a Simple and Effective Baseline (ICML 2021 · similarity 0.90)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.89)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.89)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.88)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.87)
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.