mlsanityData to 14 Oct 2026

PointTransformerX: Portable and Efficient 3D Point Cloud Processing without Sparse Algorithms

arXiv 2604.24169 ↗First posted 27 Apr 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

3D point cloud perception remains tightly coupled to custom CUDA operators for spatial operations, limiting portability and efficiency on non-NVIDIA, AMD, and embedded hardware. We introduce PointTransformerX (PTX), a fully PyTorch-native vision transformer backbone for 3D point clouds, removing all custom CUDA operators and external libraries while retaining competitive accuracy. PTX introduces 3D-GS-RoPE, a rotary positional embedding that encodes 3D spatial relationships directly in self-attention without neighborhood construction, and further replaces sparse convolutional patch embedding with a linear projection. PTX explores inference-time scaling of attention windows to improve accuracy without retraining. With a redesigned feed-forward network, PTX achieves 98.7% of PointTransformer V3's accuracy on ScanNet with 79.2% fewer parameters and executing 1.6× faster while requiring just 253 MB memory. PTX runs natively on NVIDIA GPUs, AMD GPUs (ROCm), and CPUs, providing an efficient and portable foundation for point cloud perception.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 37 of its 51 references that are in the map. First posted on arXiv 27 Apr 2026.

Builds on

37 of 51 references.

References · 37

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  3. 2016Densely Connected Convolutional Networkscited by 6,720
  4. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  5. 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 5,045
  6. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  7. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
  8. 2016Gaussian Error Linear Units (GELUs)cited by 3,502
  9. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,117
  10. 2023FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioningcited by 2,598
  11. 2019Fast Graph Representation Learning with PyTorch Geometriccited by 2,434
  12. 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,223
25 more
  1. 2020GLU Variants Improve Transformercited by 1,804
  2. 2018PointPillars: Fast Encoders for Object Detection from Point Cloudscited by 1,715
  3. 2020Point Transformercited by 1,470
  4. 20194D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networkscited by 1,208
  5. 2019KPConv: Flexible and Deformable Convolution for Point Cloudscited by 1,205
  6. 2023YaRN: Efficient Context Window Extension of Large Language Modelscited by 601
  7. 2023Point Transformer V3: Simpler, Faster, Strongercited by 589
  8. 2024FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precisioncited by 519
  9. 2021Conditional Positional Encodings for Vision Transformerscited by 406
  10. 2022Point Transformer V2: Grouped Vector Attention and Partition-based Poolingcited by 388
  11. 2023EVA-02: A Visual Representation for Neon Genesiscited by 381
  12. 2023Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision, Language, Audio, and Actioncited by 302
  13. 2024Rotary Position Embedding for Vision Transformercited by 236
  14. 20203D Object Detection with Pointformercited by 194
  15. 2023Spherical Transformer for LiDAR-based 3D Recognitioncited by 125
  16. 2023Swin3D: A Pretrained Transformer Backbone for 3D Indoor Scene Understandingcited by 113
  17. 2023OctFormer: Octree-based Transformers for 3D Point Cloudscited by 109
  18. 2022TorchSparse: Efficient Point Cloud Inference Enginecited by 73
  19. 2023FlatFormer: Flattened Window Attention for Efficient Point Cloud Transformercited by 73
  20. 2015Sparse 3D convolutional neural networkscited by 72
  21. 2024ReLU2 Wins: Discovering Efficient Activation Functions for Sparse LLMscited by 67
  22. 2024OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmentationcited by 44
  23. 2024RadarPillars: Efficient Object Detection from 4D Radar Point Cloudscited by 23
  24. 2025LitePT: Lighter Yet Stronger Point Transformercited by 17
  25. 2024Text3DAug -- Prompted Instance Augmentation for LiDAR Perceptioncited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025PointCNN++: Performant Convolution on Native Points
  2. 2021PVT: Point-Voxel Transformer for Point Cloud Learning
  3. 2025HLS4PC: A Parametrizable Framework For Accelerating Point-Based 3D Point Cloud Models on FPGA
  4. 2024CloudFixer: Test-Time Adaptation for 3D Point Clouds via Diffusion-Guided Geometric Transformation
  5. 2026Utonia: Toward One Encoder for All Point Clouds
  6. 2023PVT-SSD: Single-Stage 3D Object Detector with Point-Voxel Transformer
  7. 2025Sonata: Self-Supervised Learning of Reliable Point Representations
  8. 2025TopoDiT-3D: Topology-Aware Diffusion Transformer with Bottleneck Structure for 3D Point Cloud Generation