Self-Supervised Few-Shot Learning on Point Clouds

NeurIPS 2020  ·  Charu Sharma, Manohar Kaul ·

The increased availability of massive point clouds coupled with their utility in a wide variety of applications such as robotics, shape synthesis, and self-driving cars has attracted increased attention from both industry and academia. Recently, deep neural networks operating on labeled point clouds have shown promising results on supervised learning tasks like classification and segmentation. However, supervised learning leads to the cumbersome task of annotating the point clouds. To combat this problem, we propose two novel self-supervised pre-training tasks that encode a hierarchical partitioning of the point clouds using a cover-tree, where point cloud subsets lie within balls of varying radii at each level of the cover-tree. Furthermore, our self-supervised learning network is restricted to pre-train on the support set (comprising of scarce training examples) used to train the downstream network in a few-shot learning (FSL) setting. Finally, the fully-trained self-supervised network's point embeddings are input to the downstream task's network. We present a comprehensive empirical evaluation of our method on both downstream classification and segmentation tasks and show that supervised methods pre-trained with our self-supervised learning method significantly improve the accuracy of state-of-the-art methods. Additionally, our method also outperforms previous unsupervised methods in downstream classification tasks.

PDF Abstract NeurIPS 2020 PDF NeurIPS 2020 Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Few-Shot 3D Point Cloud Classification ModelNet40 10-way (10-shot) SSFSL+PointNet Overall Accuracy 49.15 # 21
Standard Deviation 6.1 # 21
Few-Shot 3D Point Cloud Classification ModelNet40 10-way (10-shot) SSFSL+DGCNN Overall Accuracy 48.50 # 22
Standard Deviation 5.6 # 20
Few-Shot 3D Point Cloud Classification ModelNet40 10-way (20-shot) SSFSL+ DGCNN Overall Accuracy 53.00 # 21
Standard Deviation 4.1 # 19
Few-Shot 3D Point Cloud Classification ModelNet40 10-way (20-shot) SSFSL+PointNet Overall Accuracy 50.10 # 22
Standard Deviation 5.0 # 20
Few-Shot 3D Point Cloud Classification ModelNet40 5-way (10-shot) SSFSL+DGCNN Overall Accuracy 60.0 # 22
Standard Deviation 8.9 # 19
Few-Shot 3D Point Cloud Classification ModelNet40 5-way (10-shot) SSFSL+PointNet Overall Accuracy 63.2 # 21
Standard Deviation 10.7 # 22
Few-Shot 3D Point Cloud Classification ModelNet40 5-way (20-shot) SSFSL+DGCNN Overall Accuracy 65.70 # 22
Standard Deviation 8.4 # 20
Few-Shot 3D Point Cloud Classification ModelNet40 5-way (20-shot) SSFSL+PointNet Overall Accuracy 68.90 # 20
Standard Deviation 9.4 # 21

Methods


No methods listed for this paper. Add relevant methods here