Research map / Point clouds / Point cloud classification
PointMixup: Augmentation for Point Clouds
Yunlu Chen, Vincent Tao Hu, Efstratios Gavves, Thomas Mensink, Pascal Mettes, Pengwan Yang, Cees G. M. Snoek
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 |
|---|---|---|---|
| add PointMixup (optimal assignment) interpolation (ablation) | PointNet++ · ModelNet40 full, unaligned | accuracy: 90.7 -> 91.3 | Improved |
| add manifold mixup to PointMixup (ablation) | PointNet++ · ModelNet40 reduced (20%), pre-aligned | accuracy: 87.6 -> 88.6 | Improved |
| use PointMixup instead of random assignment interpolation (ablation) | PointNet++ · ModelNet40 reduced (20%), pre-aligned | accuracy: 85.5 -> 87.6 | Improved |
| use PointMixup instead of point sampling interpolation (ablation) | PointNet++ · ModelNet40 full, pre-aligned | accuracy: 92.2 -> 92.3 | Improved |
| random assignment mixup instead of no mixup (ablation) | PointNet++ · ModelNet40 full, pre-aligned | accuracy: 91.9 -> 91.6 | Hurt |
| PointMixup versus label smoothing (0.2) (ablation) | PointNet++ · ModelNet40 reduced (20%), pre-aligned | accuracy: 87.3 -> 88.6 | Improved |
| mix input only, not labels (ablation) | PointMixup · ModelNet40 reduced (20%), pre-aligned | accuracy: 88.6 -> 86.6 | Hurt |
| mix input from same class only (ablation) | PointMixup · ModelNet40 reduced (20%), pre-aligned | accuracy: 88.6 -> 86.4 | Hurt |
| mix latent at fixed layer 1 instead of random layers (ablation) | PointMixup · ModelNet40 reduced (20%), pre-aligned | accuracy: 88.6 -> 86.9 | Hurt |
| add manifold mixup (ablation) | PointMixup · ModelNet40 noise sigma^2=0.05 | accuracy: 35.1 -> 56.5 | Improved |
| add manifold mixup (ablation) | PointMixup · ModelNet40 scale 2.0 | accuracy: 59.2 -> 72.9 | Improved |
| PointMixup without manifold mixup under Z-rotation (ablation) | PointNet++ · ModelNet40 Z-rotation [-30,30] | accuracy: 74.3 -> 70.9 | Hurt |
Later papers that compared with this method
2 later accepted papers report PointMixup as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- PSMix: Robust Point Cloud Recognition through Spectral Domain Mixing (ICML 2026 · better on 2 of 2 shared results)
- SageMix: Saliency-Guided Mixup for Point Clouds (NeurIPS 2022 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PointPatchMix: Point Cloud Mixing with Patch Scoring (AAAI 2024 · similarity 0.90)
- Regularization Strategy for Point Cloud via Rigidly Mixed Sample (CVPR 2021 · similarity 0.90)
- Point MixSwap: Attentional Point Cloud Mixing via Swapping Matched Structural Divisions (ECCV 2022 · similarity 0.90)
- Rethinking Point Cloud Data Augmentation: Topologically Consistent Deformation (ICML 2025 · similarity 0.86)
- Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2024 · similarity 0.86)
- SPU-IMR: Self-supervised Arbitrary-scale Point Cloud Upsampling via Iterative Mask-recovery Network (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.
