# Beyond Random Masking: A Dual-Stream Approach for Rotation-Invariant Point Cloud Masked Autoencoders

arXiv: https://arxiv.org/abs/2509.14975 · HTML: https://mlsanity.com/paper/2509.14975/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 37 references that are in the map.
- First posted on arXiv 18 Sep 2025.

## Builds on (most cited first, 25 of 25)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [BEiT: BERT Pre-Training of Image Transformers](https://mlsanity.com/paper/2106.08254/) (2021)
- [Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds](https://mlsanity.com/paper/1802.08219/) (2018)
- [Deep Learning for 3D Point Clouds: A Survey](https://mlsanity.com/paper/1912.12033/) (2019)
- [PointContrast: Unsupervised Pre-training for 3D Point Cloud Understanding](https://mlsanity.com/paper/2007.10985/) (2020)
- [Masked Autoencoders for Point Cloud Self-supervised Learning](https://mlsanity.com/paper/2203.06604/) (2022)
- [OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation](https://mlsanity.com/paper/2301.07525/) (2023)
- [Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training](https://mlsanity.com/paper/2205.14401/) (2022)
- [CrossPoint: Self-Supervised Cross-Modal Contrastive Learning for 3D Point Cloud Understanding](https://mlsanity.com/paper/2203.00680/) (2022)
- [Walk in the Cloud: Learning Curves for Point Clouds Shape Analysis](https://mlsanity.com/paper/2105.01288/) (2021)
- [What to Hide from Your Students: Attention-Guided Masked Image Modeling](https://mlsanity.com/paper/2203.12719/) (2022)
- [Rotation Invariant Convolutions for 3D Point Clouds Deep Learning](https://mlsanity.com/paper/1908.06297/) (2019)
- [Exploiting Edge-Oriented Reasoning for 3D Point-based Scene Graph Analysis](https://mlsanity.com/paper/2103.05558/) (2021)
- [Rotation-Invariant Local-to-Global Representation Learning for 3D Point Cloud](https://mlsanity.com/paper/2010.03318/) (2020)
- [RIConv++: Effective Rotation Invariant Convolutions for 3D Point Clouds Deep Learning](https://mlsanity.com/paper/2202.13094/) (2022)
- [Bringing Masked Autoencoders Explicit Contrastive Properties for Point Cloud Self-Supervised Learning](https://mlsanity.com/paper/2407.05862/) (2024)
- [RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learning](https://mlsanity.com/paper/2409.00353/) (2024)
- [PaRot: Patch-Wise Rotation-Invariant Network via Feature Disentanglement and Pose Restoration](https://mlsanity.com/paper/2302.02535/) (2023)
- [Rethinking Rotation Invariance with Point Cloud Registration](https://mlsanity.com/paper/2301.00149/) (2022)
- [MaskLRF: Self-supervised Pretraining via Masked Autoencoding of Local Reference Frames for Rotation-invariant 3D Point Set Analysis](https://mlsanity.com/paper/2403.00206/) (2024)
- [Point Cloud Understanding via Attention-Driven Contrastive Learning](https://mlsanity.com/paper/2411.14744/) (2024)
- [HFBRI-MAE: Handcrafted Feature Based Rotation-Invariant Masked Autoencoder for 3D Point Cloud Analysis](https://mlsanity.com/paper/2504.14132/) (2025)

## Abstract (arXiv)

Existing rotation-invariant point cloud masked autoencoders (MAE) rely on random masking strategies that overlook geometric structure and semantic coherence. Random masking treats patches independently, failing to capture spatial relationships consistent across orientations and overlooking semantic object parts that maintain identity regardless of rotation. We propose a dual-stream masking approach combining 3D Spatial Grid Masking and Progressive Semantic Masking to address these fundamental limitations. Grid masking creates structured patterns through coordinate sorting to capture geometric relationships that persist across different orientations, while semantic masking uses attention-driven clustering to discover semantically meaningful parts and maintain their coherence during masking. These complementary streams are orchestrated via curriculum learning with dynamic weighting, progressing from geometric understanding to semantic discovery. Designed as plug-and-play components, our strategies integrate into existing rotation-invariant frameworks without architectural changes, ensuring broad compatibility across different approaches. Comprehensive experiments on ModelNet40, ScanObjectNN, and OmniObject3D demonstrate consistent improvements across various rotation scenarios, showing substantial performance gains over the baseline rotation-invariant methods.