SPS-LCNN: A Significant Point Sampling-based Lightweight Convolutional Neural Network for point cloud processing

Open Webpage

Haojun Xu, Jing Bai

Published: 2023, Last Modified: 05 Nov 2025Appl. Soft Comput. 2023EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We propose a novel attention-based method called SPS.•SPS achieves the desired results with minimal overhead.•SPS can be flexibly embedded as a ”module” in other networks to improve performance.•We propose MS-SFE, which is capable of capturing rich details and global features.•SPS-LCNN is a lightweight network architecture.•SPS-LCNN achieves performance comparable to state-of-the-art methods.