Dual-Branch Center-Surrounding Contrast: Rethinking Contrastive Learning for 3D Point Clouds

This article introduces a novel approach to contrastive learning for 3D point clouds, focusing on a dual-branch architecture. The core idea revolves around contrasting center and surrounding regions within the point cloud data. The paper likely explores the effectiveness of this method in improving feature representation and downstream tasks.