Point-LN: A Lightweight Framework for Efficient Point Cloud Classification Using Non-Parametric Positional Encoding
Published in 2025 29th International Computer Conference, Computer Society of Iran (CSICC), 2025
We introduce Point-LN, a novel lightweight framework designed for efficient 3D point cloud classification. The architecture combines non-parametric building blocks such as Farthest Point Sampling (FPS), k-Nearest Neighbors (k-NN), and non-learnable positional encoding with a minimalist learnable classifier.
Point-LN delivers high classification accuracy while significantly reducing parameter count and computation time, making it well suited for real-time and resource-constrained environments. Evaluations on benchmark datasets including ModelNet40 and ScanObjectNN show that Point-LN performs competitively with state-of-the-art methods—while being much more efficient.
These results demonstrate Point-LN’s potential for scalable and deployable point cloud understanding in practical computer vision applications.