Submanifold Sparse Convolutional Networks
arXiv:1706.01307
Abstract
Convolutional network are the de-facto standard for analysing spatio-temporal data such as images, videos, 3D shapes, etc. Whilst some of this data is naturally dense (for instance, photos), many other data sources are inherently sparse. Examples include pen-strokes forming on a piece of paper, or (colored) 3D point clouds that were obtained using a LiDAR scanner or RGB-D camera. Standard "dense" implementations of convolutional networks are very inefficient when applied on such sparse data. We introduce a sparse convolutional operation tailored to processing sparse data that differs from prior work on sparse convolutional networks in that it operates strictly on submanifolds, rather than "dilating" the observation with every layer in the network. Our empirical analysis of the resulting submanifold sparse convolutional networks shows that they perform on par with state-of-the-art methods whilst requiring substantially less computation.
10 pages
Cited by in corpus (69)
- Class-balanced Grouping and Sampling for Point Cloud 3D Object Detection
- PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
- Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference
- PointCNN: Convolution On -Transformed Points
- PointConv: Deep Convolutional Networks on 3D Point Clouds
- From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
- MASC: Multi-scale Affinity with Sparse Convolution for 3D Instance Segmentation
- Prototype Adaption and Projection for Few- and Zero-shot 3D Point Cloud Semantic Segmentation
- Radar Voxel Fusion for 3D Object Detection
- Semantic Segmentation with a Sparse Convolutional Neural Network for Event Reconstruction in MicroBooNE
- Convolutional Sparse Kernel Network for Unsupervised Medical Image Analysis
- VoxelNextFusion: A Simple, Unified and Effective Voxel Fusion Framework for Multi-Modal 3D Object Detection
- PV-RCNN++: Point-Voxel Feature Set Abstraction With Local Vector Representation for 3D Object Detection
- Wire-Cell 3D Pattern Recognition Techniques for Neutrino Event Reconstruction in Large LArTPCs: Algorithm Description and Quantitative Evaluation with MicroBooNE Simulation
- ManiSkill: Generalizable Manipulation Skill Benchmark with Large-Scale Demonstrations
- Demonstration of background rejection using deep convolutional neural networks in the NEXT experiment
- QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection
- Complete & Label: A Domain Adaptation Approach to Semantic Segmentation of LiDAR Point Clouds
- MambaMIM: Pre-training Mamba with State Space Token Interpolation and its Application to Medical Image Segmentation
- Spherical Kernel for Efficient Graph Convolution on 3D Point Clouds
- Associate-3Ddet: Perceptual-to-Conceptual Association for 3D Point Cloud Object Detection
- An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds
- Graph neural network for 3D classification of ambiguities and optical crosstalk in scintillator-based neutrino detectors
- Task-Aware Sampling Layer for Point-Wise Analysis
- SG-NN: Sparse Generative Neural Networks for Self-Supervised Scene Completion of RGB-D Scans
- A Survey on Deep Domain Adaptation and Tiny Object Detection Challenges, Techniques and Datasets
- Deep Point Cloud Reconstruction
- Point Proposal Network for Reconstructing 3D Particle Endpoints with Sub-Pixel Precision in Liquid Argon Time Projection Chambers
- PV-RCNN: The Top-Performing LiDAR-only Solutions for 3D Detection / 3D Tracking / Domain Adaptation of Waymo Open Dataset Challenges
- RSN: Range Sparse Net for Efficient, Accurate LiDAR 3D Object Detection
- Deep Residual 3D U-Net for Joint Segmentation and Texture Classification of Nodules in Lung
- Real-time 3D object proposal generation and classification under limited processing resources
- RevealNet: Seeing Behind Objects in RGB-D Scans
- Streaming Object Detection for 3-D Point Clouds
- Real-Time Anchor-Free Single-Stage 3D Detection with IoU-Awareness
- Equivariant Maps for Hierarchical Structures
- Efficient Semantic Scene Completion Network with Spatial Group Convolution
- Unsupervised Data Imputation via Variational Inference of Deep Subspaces
- NeSF: Neural Semantic Fields for Generalizable Semantic Segmentation of 3D Scenes
- Deep Octree-based CNNs with Output-Guided Skip Connections for 3D Shape and Scene Completion
- PFA: Privacy-preserving Federated Adaptation for Effective Model Personalization
- PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features
- MLOD: Awareness of Extrinsic Perturbation in Multi-LiDAR 3D Object Detection for Autonomous Driving
- Adversarial methods to reduce simulation bias in neutrino interaction event filtering at Liquid Argon Time Projection Chambers
- Multimodal End-to-End Sparse Model for Emotion Recognition
- DV-ConvNet: Fully Convolutional Deep Learning on Point Clouds with Dynamic Voxelization and 3D Group Convolution
- Accelerated CNN Training Through Gradient Approximation
- Efficient and accurate object detection with simultaneous classification and tracking
- PointMixer: MLP-Mixer for Point Cloud Understanding
- Image-Based Jet Analysis
- Structure Information is the Key: Self-Attention RoI Feature Extractor in 3D Object Detection
- TrackFormers: In Search of Transformer-Based Particle Tracking for the High-Luminosity LHC Era
- 3D-FCT: Simultaneous 3D Object Detection and Tracking Using Feature Correlation
- Cirrus: A Long-range Bi-pattern LiDAR Dataset
- SparseDet: Towards End-to-End 3D Object Detection
- GIPSO: Geometrically Informed Propagation for Online Adaptation in 3D LiDAR Segmentation
- CIRCLE: Convolutional Implicit Reconstruction and Completion for Large-scale Indoor Scene
- To the Point: Efficient 3D Object Detection in the Range Image with Graph Convolution Kernels
- Video-based Point Cloud Compression Artifact Removal
- DV-Det: Efficient 3D Point Cloud Object Detection with Dynamic Voxelization
- Deep Convolution for Irregularly Sampled Temporal Point Clouds
- GTT-Net: Learned Generalized Trajectory Triangulation
- AccSS3D: Accelerator for Spatially Sparse 3D DNNs
- R-AGNO-RPN: A LIDAR-Camera Region Deep Network for Resolution-Agnostic Detection
- Boundary-Aware Dense Feature Indicator for Single-Stage 3D Object Detection from Point Clouds
- RoIFusion: 3D Object Detection from LiDAR and Vision
- Indoor Scene Recognition in 3D
- Potential Convolution: Embedding Point Clouds into Potential Fields
- Scaling the training of particle classification on simulated MicroBooNE events to multiple GPUs