Going Deeper with Lean Point Networks
arXiv:1907.00960
Abstract
In this work we introduce Lean Point Networks (LPNs) to train deeper and more accurate point processing networks by relying on three novel point processing blocks that improve memory consumption, inference time, and accuracy: a convolution-type block for point sets that blends neighborhood information in a memory-efficient manner; a crosslink block that efficiently shares information across low- and high-resolution processing branches; and a multiresolution point cloud processing block for faster diffusion of information. By combining these blocks, we design wider and deeper point-based architectures. We report systematic accuracy and memory consumption improvements on multiple publicly available segmentation tasks by using our generic modules as drop-in replacements for the blocks of multiple architectures (PointNet++, DGCNN, SpiderNet, PointCNN).
arXiv comment16 pages, 11 figures, 9 tables
References in corpus (12)
- Densely Connected Convolutional Networks
- ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes
- Training Deep Nets with Sublinear Memory Cost
- Aggregated Residual Transformations for Deep Neural Networks
- PCPNET: Learning Local Shape Properties from Raw Point Clouds
- Matterport3D: Learning from RGB-D Data in Indoor Environments
- Dynamic Graph CNN for Learning on Point Clouds
- The Reversible Residual Network: Backpropagation Without Storing Activations
- DeepGCNs: Can GCNs Go as Deep as CNNs?
- Relation-Shape Convolutional Neural Network for Point Cloud Analysis
- PU-Net: Point Cloud Upsampling Network
- SPLATNet: Sparse Lattice Networks for Point Cloud Processing