Weakly Supervised 3D Instance Segmentation without Instance-level Annotations

TL;DR

This paper introduces a novel weakly-supervised 3D instance segmentation method that relies solely on semantic labels, eliminating the need for instance-level annotations and significantly reducing labeling effort.

Contribution

It is the first approach to perform 3D instance segmentation using only semantic labels, employing pseudo labels and shape-aware signals for training.

Findings

Achieves comparable results to fully supervised methods.

Requires only 0.02% of total points for semantic annotations.

Reduces annotation cost for 3D instance segmentation.

Abstract

3D semantic scene understanding tasks have achieved great success with the emergence of deep learning, but often require a huge amount of manually annotated training data. To alleviate the annotation cost, we propose the first weakly-supervised 3D instance segmentation method that only requires categorical semantic labels as supervision, and we do not need instance-level labels. The required semantic annotations can be either dense or extreme sparse (e.g. 0.02% of total points). Even without having any instance-related ground-truth, we design an approach to break point clouds into raw fragments and find the most confident samples for learning instance centroids. Furthermore, we construct a recomposed dataset using pseudo instances, which is used to learn our defined multilevel shape-aware objectness signal. An asymmetrical object inference algorithm is followed to process core points and boundary points with different strategies, and generate high-quality pseudo instance labels to guide iterative training. Experiments demonstrate that our method can achieve comparable results with recent fully supervised methods. By generating pseudo instance labels from categorical semantic labels, our designed approach can also assist existing methods for learning 3D instance segmentation at reduced annotation cost.