Unsupervised Multi-Task Feature Learning on Point Clouds

TL;DR

This paper presents an unsupervised multi-task learning framework for point cloud feature extraction, combining clustering, reconstruction, and classification tasks to improve shape understanding without labeled data.

Contribution

It introduces a novel multi-task, multi-scale graph-based encoder that jointly learns point and shape features in an unsupervised manner, outperforming previous models.

Findings

Achieves 89.1% accuracy on ModelNet40 classification.

Attains 68.2 mIoU on ShapeNet segmentation.

Outperforms prior state-of-the-art unsupervised models.

Abstract

We introduce an unsupervised multi-task model to jointly learn point and shape features on point clouds. We define three unsupervised tasks including clustering, reconstruction, and self-supervised classification to train a multi-scale graph-based encoder. We evaluate our model on shape classification and segmentation benchmarks. The results suggest that it outperforms prior state-of-the-art unsupervised models: In the ModelNet40 classification task, it achieves an accuracy of 89.1% and in ShapeNet segmentation task, it achieves an mIoU of 68.2 and accuracy of 88.6%.