DAAL-WS: A weakly-supervised method integrated with data augmentation and active learning strategies for MLS point cloud semantic segmentation

Xiangda Lei, Haiyan Guan, Lingfei Ma, Jiacheng Liu, Yogntao Yu, Lanying Wang, Zhen Dong, Huan Ni, Jonathan Li

Abstract: Highlights•An EC-Mix3D data augmentation strategy is designed to augment the training data.•A PLEP active learning strategy is presented to effectively reduce label redundancy.•The proposed DAAL-WS model achieves competitive accuracy using only sparse labels.