Abstract:Point cloud segmentation is a key problem in 3D content understanding. The existing methods based on deep neural network for point cloud segmentation mainly train the network in a supervised fashion, which heavily rely on a large amount of high-quality manual-labeled training point clouds. However, it is very tedious and time-consuming to manually assign part labels for each point in point clouds. Meanwhile, a lot of unlabeled point clouds can easily be obtained from 3D scanners, Internet or reconstruction. Therefore, we introduce self-training to utilize these unlabeled point clouds. So the proposed semi-supervised point cloud segmentation method can employ both labeled point clouds and unlabeled point clouds for training. Moreover, in order to make better use of unlabeled point clouds, the adopted adversarial architecture proposes confidence discrimination of label prediction for unlabeled point clouds. Thus, pseudo labels on unlabeled point clouds with higher reliability can be picked out to participate the network training, which further improves segmentation performance. The experiments show that the proposed method can make full use of the unlabeled point clouds in training. In addition, segmentation performance improves by self-training with label confidence prediction. Point cloud segmentation is a key problem in 3D content understanding. The existing methods based on deep neural network for point cloud segmentation mainly train the network in a supervised fashion, which heavily rely on a large amount of high-quality manual-labeled training point clouds. However, it is very tedious and time-consuming to manually assign part labels for each point in point clouds. Meanwhile, a lot of unlabeled point clouds can easily be obtained from 3D scanners, Internet or reconstruction. Therefore, we introduce self-training to utilize these unlabeled point clouds. So the proposed semi-supervised point cloud segmentation method can employ both labeled point clouds and unlabeled point clouds for training. Moreover, in order to make better use of unlabeled point clouds, the adopted adversarial architecture proposes confidence discrimination of label prediction for unlabeled point clouds. Thus, pseudo labels on unlabeled point clouds with higher reliability can be picked out to participate the network training, which further improves segmentation performance. The experiments show that the proposed method can make full use of the unlabeled point clouds in training. In addition, segmentation performance improves by self-training with label confidence prediction. (c) 2021 Elsevier B.V. All rights reserved.

Weakly-Supervised Point Cloud Segmentation Combining Pseudo Label Generation and Noise Label Learning

Superpoint-guided Semi-supervised Semantic Segmentation of 3D Point Clouds

Weakly Supervised 3D Segmentation Via Receptive-Driven Pseudo Label Consistency and Structural Consistency.

WEAKLY SUPERVISED PSEUDO-LABEL ASSISTED LEARNING FOR ALS POINT CLOUD SEMANTIC SEGMENTATION

SSPC-Net: Semi-supervised Semantic 3D Point Cloud Segmentation Network

Weakly Supervised Semantic Segmentation for Large-Scale Point Cloud

Fully Data-Driven Pseudo Label Estimation for Pointly-Supervised Panoptic Segmentation

Weakly-Supervised Point Cloud Semantic Segmentation Based on Dilated Region

Learning from Weak and Noisy Labels for Semantic Segmentation.

Semi-supervised point cloud segmentation using self-training with label confidence prediction

Weakly Supervised 3D Instance Segmentation without Instance-level Annotations

Weakly Supervised Point Cloud Segmentation Via Deep Morphological Semantic Information Embedding

Joint Global and Dynamic Pseudo Labeling for Semi-Supervised Point Cloud Sequence Segmentation

Weakly supervised point cloud semantic segmentation with the fusion of heterogeneous network features

Image Understands Point Cloud: Weakly Supervised 3D Semantic Segmentation via Association Learning

Less is More: Label Recommendation for Weakly Supervised Point Cloud Semantic Segmentation

Dense Supervision Propagation for Weakly Supervised Semantic Segmentation on 3D Point Clouds

Point Cloud Instance Segmentation with Inaccurate Bounding-Box Annotations

Weakly Supervised Semantic Segmentation of Point Cloud Scenes Using Boundary-based Feature Aggregation

SegGroup: Seg-Level Supervision for 3D Instance and Semantic Segmentation

Reliability-Adaptive Consistency Regularization for Weakly-Supervised Point Cloud Segmentation