Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Point-LN: A Lightweight Framework for Efficient Point Cloud Classification Using Non-Parametric Positional Encoding

Marzieh Mohammadi, Amir Salarpour, Pedram MohajerAnsari

TL;DR

Point-LN tackles efficient 3D point cloud classification by merging non-parametric positional encodings with a lightweight classifier. It leverages FPS, KNN, and non-learnable encodings (TPE and GPE) within a four-stage feature encoder to maintain high accuracy with roughly 0.8M parameters. On ModelNet40 and ScanObjectNN, Point-LN delivers competitive results while significantly reducing model size compared to state-of-the-art parametric methods, highlighting strong efficiency and scalability. This approach enables real-time 3D recognition on resource-constrained devices, with potential extensions to segmentation and object detection.

Abstract

We introduce Point-LN, a novel lightweight framework engineered for efficient 3D point cloud classification. Point-LN integrates essential non-parametric components-such as Farthest Point Sampling (FPS), k-Nearest Neighbors (k-NN), and non-learnable positional encoding-with a streamlined learnable classifier that significantly enhances classification accuracy while maintaining a minimal parameter footprint. This hybrid architecture ensures low computational costs and rapid inference speeds, making Point-LN ideal for real-time and resource-constrained applications. Comprehensive evaluations on benchmark datasets, including ModelNet40 and ScanObjectNN, demonstrate that Point-LN achieves competitive performance compared to state-of-the-art methods, all while offering exceptional efficiency. These results establish Point-LN as a robust and scalable solution for diverse point cloud classification tasks, highlighting its potential for widespread adoption in various computer vision applications.

Point-LN: A Lightweight Framework for Efficient Point Cloud Classification Using Non-Parametric Positional Encoding

TL;DR

Point-LN tackles efficient 3D point cloud classification by merging non-parametric positional encodings with a lightweight classifier. It leverages FPS, KNN, and non-learnable encodings (TPE and GPE) within a four-stage feature encoder to maintain high accuracy with roughly 0.8M parameters. On ModelNet40 and ScanObjectNN, Point-LN delivers competitive results while significantly reducing model size compared to state-of-the-art parametric methods, highlighting strong efficiency and scalability. This approach enables real-time 3D recognition on resource-constrained devices, with potential extensions to segmentation and object detection.

Abstract

We introduce Point-LN, a novel lightweight framework engineered for efficient 3D point cloud classification. Point-LN integrates essential non-parametric components-such as Farthest Point Sampling (FPS), k-Nearest Neighbors (k-NN), and non-learnable positional encoding-with a streamlined learnable classifier that significantly enhances classification accuracy while maintaining a minimal parameter footprint. This hybrid architecture ensures low computational costs and rapid inference speeds, making Point-LN ideal for real-time and resource-constrained applications. Comprehensive evaluations on benchmark datasets, including ModelNet40 and ScanObjectNN, demonstrate that Point-LN achieves competitive performance compared to state-of-the-art methods, all while offering exceptional efficiency. These results establish Point-LN as a robust and scalable solution for diverse point cloud classification tasks, highlighting its potential for widespread adoption in various computer vision applications.
Paper Structure (17 sections, 4 equations, 2 figures, 2 tables)

This paper contains 17 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Projection-based methods
  4. Point-based Methods
  5. Parametric Methods
  6. Non-parametric Methods
  7. Method
  8. Background
  9. Positional Encoding
  10. Lightweight Feature Encoder
  11. Classifier Architecture
  12. Experiments
  13. Experimental Setup
  14. Dataset Details
  15. Shape classification on ModelNet40
  16. ...and 2 more sections

Figures (2)

  • Figure 1: Illustration of Point-LN for Point Cloud Classification.
  • Figure 2: Overview of the proposed network architecture: the feature encoder extracts high-dimensional representations from raw point clouds, and the classifier maps these features to the target label space.