Efficient Point Cloud Processing with High-Dimensional Positional Encoding and Non-Local MLPs

2026-03-04

AI Summary

This work addresses the high computational cost of conventional MLP-based point cloud models, which stems from complex local operations and hinders the balance between performance and efficiency. To overcome this limitation, the authors propose ABS-REF, a two-stage paradigm that explicitly models the intrinsic geometric structure of point clouds through High-dimensional Positional Encoding (HPE) and replaces local operations with non-local MLPs to effectively integrate both local and global contextual information. The resulting architecture, HPENet, achieves significant performance gains over PointNeXt across seven benchmark datasets—including ScanObjectNN—while simultaneously reducing computational complexity: FLOPs are cut to 21.5%–50% of the baseline, accompanied by consistent improvements in multiple metrics such as mean accuracy (mAcc) and mean Intersection over Union (mIoU).

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Large Vision Models
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Multi-Layer Perceptron (MLP) models are the foundation of contemporary point cloud processing. However, their complex network architectures obscure the source of their strength and limit the application of these models. In this article, we develop a two-stage abstraction and refinement (ABS-REF) view for modular feature extraction in point cloud processing. This view elucidates that whereas the early models focused on ABS stages, the more recent techniques devise sophisticated REF stages to attain performance advantages. Then, we propose a High-dimensional Positional Encoding (HPE) module to explicitly utilize intrinsic positional information, extending the ``positional encoding'' concept from Transformer literature. HPE can be readily deployed in MLP-based architectures and is compatible with transformer-based methods. Within our ABS-REF view, we rethink local aggregation in MLP-based methods and propose replacing time-consuming local MLP operations, which are used to capture local relationships among neighbors. Instead, we use non-local MLPs for efficient non-local information updates, combined with the proposed HPE for effective local information representation. We leverage our modules to develop HPENets, a suite of MLP networks that follow the ABS-REF paradigm, incorporating a scalable HPE-based REF stage. Extensive experiments on seven public datasets across four different tasks show that HPENets deliver a strong balance between efficiency and effectiveness. Notably, HPENet surpasses PointNeXt, a strong MLP-based counterpart, by 1.1% mAcc, 4.0% mIoU, 1.8% mIoU, and 0.2% Cls. mIoU, with only 50.0%, 21.5%, 23.1%, 44.4% of FLOPs on ScanObjectNN, S3DIS, ScanNet, and ShapeNetPart, respectively. Source code is available at https://github.com/zouyanmei/HPENet_v2.git.

Problem

• Point Cloud Processing
• MLP Models
• Efficiency
• Positional Encoding
• Local Aggregation

Innovation

• High-dimensional Positional Encoding
• Non-Local MLPs
• ABS-REF Framework
• Point Cloud Processing
• Efficient Feature Extraction