SO-Net: Self-Organizing Network for Point Cloud Analysis

Li, Jiaxin; Chen, Ben M.; Lee, Gim Hee

doi:10.1109/cvpr.2018.00979

Cited by 896 publications

(572 citation statements)

References 37 publications

Supporting

Mentioning

569

Contrasting

Unclassified

Order By: Relevance

“…PPF-FoldNet [13] projects the points into point pair feature (PPF) space and then applies a PointNet encoder and a Fold-ingNet decoder to reconstruct that space. AtlasNet [26] extends the FoldingNet to multiple grid patches whereas SO-Net [39] aggregates the point features into SOM node features to encode the spatial distributions. PointCapsNet [96] introduces an autoencoder based on dynamic routing to extract latent capsules and a few MLPs that generate multiple point patches from the latent capsules with distinct grids.…”

Section: Deep Learning On Point Cloudsmentioning

confidence: 99%

Unsupervised Multi-Task Feature Learning on Point Clouds

Hassani

Haley

2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

183

117

View full text Add to dashboard Cite

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%.

show abstract

Section: Deep Learning On Point Cloudsmentioning

confidence: 99%

Unsupervised Multi-Task Feature Learning on Point Clouds

Hassani

Haley

2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

183

117

View full text Add to dashboard Cite

show abstract

“…Following previous studies Li et al (2018a); Qi et al (2017b), the relative coordinates are adopted in LRC-Net. Before feeding points inside each local region R j into the PointNet layer, a relative coordinate system of the centroid p j is built by a simple operation: p l = p l − p j , where l is the index of points in the local region R j .…”

Section: Area Feature Extractionmentioning

confidence: 99%

LRC-Net: Learning discriminative features on point clouds by encoding local region contexts

Liu

Han

Hong

et al. 2020

Computer Aided Geometric Design

View full text Add to dashboard Cite

Learning discriminative feature directly on point clouds is still challenging in the understanding of 3D shapes. Recent methods usually partition point clouds into local region sets, and then extract the local region features with fixed-size CNN or MLP, and finally aggregate all individual local features into a global feature using simple max pooling. However, due to the irregularity and sparsity in sampled point clouds, it is hard to encode the fine-grained geometry of local regions and their spatial relationships when only using the fixed-size filters and individual local feature integration, which limit the ability to learn discriminative features. To address this issue, we present a novel Local-Region-Context Network (LRC-Net), to learn discriminative features on point clouds by encoding the fine-grained contexts inside and among local regions simultaneously. LRC-Net consists of two main modules. The first module, named intra-region context encoding, is designed for capturing the geometric correlation inside each local region by novel variable-size convolution filter. The second module, named inter-region context encoding, is proposed for integrating the spatial relationships among local regions based on spatial similarity measures. Experimental results show that LRC-Net is competitive with state-of-the-art methods in shape classification and shape segmentation applications.

show abstract

“…Application Parameter Data Model (Image set) (feature) (simulations) 36 fluids position KNN (SOM) microscope 38 cellular position HF dynamics microscope 39 self-assembly cluster DM (distance) video 40 surveillance object CNN+RNN cryo-EM 41 macromolecules object deep CNN particle high-energy track LSTM+CNN detector 42 physics holograms 43 colloidal 3D position CC,CNN science MNIST 44 computer position SO-Net science cloud…”

Section: Instrumentmentioning

confidence: 99%

Microparticle cloud imaging and tracking for data-driven plasma science

et al. 2020

View full text Add to dashboard Cite

Large data sets give rise to the 'fourth paradigm' of scientific discovery and technology development, extending other approaches based on human intuition, fundamental laws of physics, statistics and intense computation. Both experimental and simulation data are growing explosively in plasma science and technology, motivating data-driven discoveries and inventions, which are currently in infancy. Here we describe recent progress in microparticle cloud imaging and tracking (mCIT, µCIT) for laboratory plasma experiments. Three types of microparticle clouds are described: from exploding wires, in dusty plasmas and in atmospheric plasmas. The experimental data sets are obtained with one or more imaging cameras at a rate up to 100k frames per second (fps). Analyses of the time-dependent microparticle trajectories give time-dependent two-dimensional or three-dimensional information about the particle motion and ambient environment. The massive image and particle track data motivate development of machine-learning (ML) techniques for information extraction. A physicsconstrained motion tracker, a Kohonen neural network (KNN) or self-organizing map (SOM), the feature tracking kit (FTK), and U-Net are described and compared with each other for particle tracking using the datasets. Particle density and signal-to-noise ratio have been identified as two important factors that affect the tracking accuracy. Fast Fourier transform (FFT) is used to reveal how U-Net, a deep convolutional neural network (CNN) developed for non-plasma applications, achieves the improvements for noisy scenes. The fitting parameters for a simple polynomial track model are found to group into clusters that reveal the geometry information about the camera setup. The mCIT or µCIT techniques, when enhanced with data models, can be used to study the microparticle-or Debye-length scale plasma physics. The datasets are also available for ML code development and comparisons of algorithms.

show abstract

SO-Net: Self-Organizing Network for Point Cloud Analysis

Cited by 896 publications

References 37 publications

Unsupervised Multi-Task Feature Learning on Point Clouds

Unsupervised Multi-Task Feature Learning on Point Clouds

LRC-Net: Learning discriminative features on point clouds by encoding local region contexts

Microparticle cloud imaging and tracking for data-driven plasma science

Contact Info

Product

Resources

About