Research map / Point clouds / Point cloud classification
Efficient Point Cloud Processing With High-Dimensional Positional Encoding and Non-Local MLPs
Yanmei Zou, Hongshan Yu, Yaonan Wang, Zhengeng Yang, Xieyuanli Chen, Kailun Yang, Naveed Akhtar
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 HPE to the ABS stage only (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 63.9 -> 70.0 | Improved |
| add HPE to the REF stage only (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 63.9 -> 70.7 | Improved |
| add HPE to both ABS and REF stages (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 63.9 -> 72.2 | Improved |
| add Backward Fusion Module (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 72.2 -> 72.3 | No clear effect |
| use high-dimensional HPE instead of 3-dim PE (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 70.9 -> 72.2 | Improved |
| use default HPE instead of multiplicative HPE (mul) (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 71.0 -> 72.2 | Improved |
| use default HPE instead of absolute-coordinate variant (abs) (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 54.6 -> 72.2 | Improved |
| increase HPE_MLP dimension from 3 to C/4 (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 70.9 -> 72.2 | Improved |
| increase HPE_MLP dimension from C/4 to C (ablation) | HPENet V2-XL · S3DIS Area-5 | mIoU: 72.2 -> 71.8 | Hurt |
| use PreConv instead of Conv in ABS stage (efficient aggregation) (ablation) | Conv/Conv model · S3DIS Area-5 | mIoU: 66.6 -> 65.6 | Hurt |
| add positional encoding to PreConv (ablation) | PreConv/PreConv · S3DIS Area-5 | mIoU: 58.8 -> 66.7 | Improved |
| replace PE with HPE in hybrid PreConv model (ablation) | PreConv+Conv first ABS · S3DIS Area-5 | mIoU: 67.1 -> 68.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointNeXt-XL — PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies (NeurIPS 2022) · better on 2 of 2 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
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.96)
- Point Cloud Mamba: Point Cloud Learning via State Space Model (AAAI 2025 · similarity 0.89)
- Point Deformable Network with Enhanced Normal Embedding for Point Cloud Analysis (AAAI 2024 · similarity 0.89)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.88)
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.88)
- LCM: Locally Constrained Compact Point Cloud Model for Masked Point Modeling (NeurIPS 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.
