Filters








46 Hits in 2.6 sec

PointFlow: 3D Point Cloud Generation with Continuous Normalizing Flows [article]

Guandao Yang, Xun Huang, Zekun Hao, Ming-Yu Liu, Serge Belongie, Bharath Hariharan
2019 arXiv   pre-print
Our generative model, named PointFlow, learns each level of the distribution with a continuous normalizing flow.  ...  Empirically, we demonstrate that PointFlow achieves state-of-the-art performance in point cloud generation.  ...  Conclusion and future works In this paper, we propose PointFlow, a generative model for point clouds consisting of two levels of continuous normalizing flows trained with variational inference.  ... 
arXiv:1906.12320v3 fatcat:u2ilsq2fsfb77m653oeatt4674

HyperFlow: Representing 3D Objects as Surfaces [article]

Przemysław Spurek, Maciej Zięba, Jacek Tabor, Tomasz Trzciński
2020 arXiv   pre-print
To that end, we devise a generative model that uses a hypernetwork to return the weights of a Continuous Normalizing Flows (CNF) target network.  ...  In this work, we present HyperFlow - a novel generative model that leverages hypernetworks to create continuous 3D object representations in a form of lightweight surfaces (meshes), directly out of point  ...  Flow [17] target network and generate 3D point clouds together with its mesh-based representation.  ... 
arXiv:2006.08710v1 fatcat:wimfdur35ra2zp7ap53i5zxfyq

Discrete Point Flow Networks for Efficient Point Cloud Generation [article]

Roman Klokov, Edmond Boyer, Jakob Verbeek
2020 arXiv   pre-print
We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation.  ...  In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed.  ...  . , x n } be a point cloud with x i ∈ IR d , where d = 3 for point clouds for 3D shapes. The dimension d may be larger in some cases, e.g. d = 6 when modeling 3D points equipped with surface normals.  ... 
arXiv:2007.10170v1 fatcat:y2ibl7ijffgkrjowo2aasa4zcu

General hypernetwork framework for creating 3D point clouds

Przemysaw Spurek, Maciej Zieba, Jacek Tabor, Tomasz Trzcinski
2021 IEEE Transactions on Pattern Analysis and Machine Intelligence  
In this work, we propose a novel method for generating 3D point clouds that leverages properties of hypernetworks.  ...  As a consequence, a particular 3D shape can be generated using point-by-point sampling from the prior distribution and transforming sampled points with the target network.  ...  HYPERFLOW: HYPERNETWORK AND CONTINU-OUS NORMALIZING FLOWS FOR GENERATING 3D POINT CLOUDS In this section, we present our general framework for creating 3D point clouds, together with their mesh-based representations  ... 
doi:10.1109/tpami.2021.3131131 pmid:34826294 fatcat:nhmv2zmqcrhghadtxpjyhek3zi

Hypernetwork approach to generating point clouds [article]

Przemysław Spurek, Sebastian Winczowski, Jacek Tabor, Maciej Zamorski, Maciej Zięba, Tomasz Trzciński
2020 arXiv   pre-print
In this work, we propose a novel method for generating 3D point clouds that leverage properties of hyper networks.  ...  The proposed architecture allows finding mesh-based representation of 3D objects in a generative manner while providing point clouds en pair in quality with the state-of-the-art methods.  ...  PointFlow uses continuous normalizing flow Grathwohl et al., 2018) for both of these tasks.  ... 
arXiv:2003.00802v2 fatcat:kcbf2le6efghxddsrlwxspulbm

Representing Point Clouds with Generative Conditional Invertible Flow Networks [article]

Michał Stypułkowski, Kacper Kania, Maciej Zamorski, Maciej Zięba, Tomasz Trzciński, Jan Chorowski
2020 arXiv   pre-print
Our method leverages generative invertible flow networks to learn embeddings as well as to generate point clouds.  ...  parameters with the exception of a small, object-specific embedding vector.  ...  One of the most recent methods, PointFlow [23] , utilizes VAE, together with continuous normalizing flows [3] , to generate 3D point clouds.  ... 
arXiv:2010.11087v1 fatcat:5az7wtkzsrhdbaky4byduvturi

SoftFlow: Probabilistic Framework for Normalizing Flow on Manifolds [article]

Hyeongju Kim, Hyeonseung Lee, Woo Hyun Kang, Joun Yeop Lee, Nam Soo Kim
2020 arXiv   pre-print
The proposed model for 3D point clouds, namely SoftPointFlow, can estimate the distribution of various shapes more accurately and achieves state-of-the-art performance in point cloud generation.  ...  Furthermore, we apply the proposed framework to 3D point clouds to alleviate the difficulty of forming thin structures for flow-based models.  ...  Furthermore, we also propose SoftPointFlow for 3D point cloud generation which relieves the difficulty of forming thin structures.  ... 
arXiv:2006.04604v4 fatcat:qa4nn4phtbgfbmqq4uuacnacr4

ChartPointFlow for Topology-Aware 3D Point Cloud Generation [article]

Takumi Kimura, Takashi Matsubara, Kuniaki Uehara
2020 arXiv   pre-print
In this paper, we propose ChartPointFlow, which is a flow-based generative model with multiple latent labels.  ...  By maximizing the mutual information, a map conditioned by a label is assigned to a continuous subset of a given point cloud, like a chart of a manifold.  ...  SoftFlow did not employ FFJORD for 3D point cloud generation. A flow-based generative model always learns a continuous deformation.  ... 
arXiv:2012.02346v1 fatcat:eh3z5lpz7vboppmk4omvks3yjq

Go with the Flows: Mixtures of Normalizing Flows for Point Cloud Generation and Reconstruction [article]

Janis Postels, Mengya Liu, Riccardo Spezialetti, Luc Van Gool, Federico Tombari
2021 arXiv   pre-print
Recently normalizing flows (NFs) have demonstrated state-of-the-art performance on modeling 3D point clouds while allowing sampling with arbitrary resolution at inference time.  ...  By instantiating each mixture component with a comparatively small NF we generate point clouds with improved details compared to single-flow-based models while using fewer parameters and considerably reducing  ...  [44] generates the weights of the continuous NF using a hypernetwork paired with a spherical log-normal base distribution achieving similar results as PointFlow.  ... 
arXiv:2106.03135v3 fatcat:uontquco3jbg7dvf62z5imcd4m

Progressive Point Cloud Deconvolution Generation Network [article]

Le Hui, Rui Xu, Jin Xie, Jianjun Qian, Jian Yang
2020 arXiv   pre-print
In this paper, we propose an effective point cloud generation method, which can generate multi-resolution point clouds of the same shape from a latent vector.  ...  In order to keep the shapes of different resolutions of point clouds consistent, we propose a shape-preserving adversarial loss to train the point cloud deconvolution generation network.  ...  .: Pointflow: 3d point cloud generation with continuous normalizing flows. In: ICCV (2019) 45. Yang, Y., Feng, C., Shen, Y., Tian, D.: Foldingnet: Point cloud auto-encoder via deep grid deformation.  ... 
arXiv:2007.05361v1 fatcat:ads6bb55bzdflbsbavgnz4v3ui

Visibility-Aware Point-Based Multi-View Stereo Network

Rui Chen, Songfang Han, Jing Xu, hao su
2020 IEEE Transactions on Pattern Analysis and Machine Intelligence  
Our network leverages 3D geometry priors and 2D texture information jointly and effectively by fusing them into a feature-augmented point cloud, and processes the point cloud to estimate the 3D flow for  ...  We first generate a coarse depth map, convert it into a point cloud and refine the point cloud iteratively by estimating the residual between the depth of the current iteration and that of the ground truth  ...  Top: Whole point cloud. Bottom: Visualization of normals in zoomed local area. Our VA-Point-MVSNet generates detailed point clouds with more high-frequency components than MVSNet.  ... 
doi:10.1109/tpami.2020.2988729 pmid:32324542 fatcat:dwdbcnncxrb2jdijvs63ckdndq

CaSPR: Learning Canonical Spatiotemporal Point Cloud Representations [article]

Davis Rempe, Tolga Birdal, Yongheng Zhao, Zan Gojcic, Srinath Sridhar, Leonidas J. Guibas
2020 arXiv   pre-print
normalizing flows.  ...  Different from previous work, CaSPR learns representations that support spacetime continuity, are robust to variable and irregularly spacetime-sampled point clouds, and generalize to unseen object instances  ...  The authors thank Michael Niemeyer for providing the code and shape models used to generate the warping cars dataset.  ... 
arXiv:2008.02792v2 fatcat:xxn7tf5r3vf35exzpajhqzeazi

Diffusion Probabilistic Models for 3D Point Cloud Generation [article]

Shitong Luo, Wei Hu
2021 arXiv   pre-print
We present a probabilistic model for point cloud generation, which is fundamental for various 3D vision tasks such as shape completion, upsampling, synthesis and data augmentation.  ...  Inspired by the diffusion process in non-equilibrium thermodynamics, we view points in point clouds as particles in a thermodynamic system in contact with a heat bath, which diffuse from the original distribution  ...  of point clouds.  ... 
arXiv:2103.01458v2 fatcat:ouxzyxpg6nfjffldnnkui425wy

Learning Gradient Fields for Shape Generation [article]

Ruojin Cai, Guandao Yang, Hadar Averbuch-Elor, Zekun Hao, Serge Belongie, Noah Snavely, Bharath Hariharan
2020 arXiv   pre-print
A point cloud can be viewed as samples from a distribution of 3D points whose density is concentrated near the surface of the shape.  ...  Point cloud generation thus amounts to moving randomly sampled points to high-density areas.  ...  PointFlow [64] applies normalizing flow [46] to model such distribution, so sampling points amounts to moving them to the surface according to a learned vector field.  ... 
arXiv:2008.06520v2 fatcat:pot37vogyrgczbe2eelpnyhaba

Generative PointNet: Deep Energy-Based Learning on Unordered Point Sets for 3D Generation, Reconstruction and Classification [article]

Jianwen Xie, Yifei Xu, Zilong Zheng, Song-Chun Zhu, Ying Nian Wu
2021 arXiv   pre-print
Furthermore, we can learn a short-run MCMC toward the energy-based model as a flow-like generator for point cloud reconstruction and interpolation.  ...  Unlike most point cloud generators that rely on hand-crafted distance metrics, our model does not require any hand-crafted distance metric for the point cloud generation, because it synthesizes point clouds  ...  [50] studies point cloud generation using continuous normalizing flows trained with variational inference. Our paper learns an EBM of point clouds via MCMC-based MLE.  ... 
arXiv:2004.01301v2 fatcat:adqvmj7khzhcnpsln22tud2shi
« Previous Showing results 1 — 15 out of 46 results