The Experts below are selected from a list of 81975 Experts worldwide ranked by ideXlab platform
Bharath Hariharan - One of the best experts on this subject based on the ideXlab platform.
-
Pointflow 3d Point cloud generation with continuous normalizing flows
International Conference on Computer Vision, 2019Co-Authors: Guandao Yang, Xun Huang, Serge Belongie, Bharath HariharanAbstract:As 3D Point clouds become the representation of choice for multiple vision and graphics applications, the ability to synthesize or reconstruct high-resolution, high-fidelity Point clouds becomes crucial. Despite the recent success of deep learning models in discriminative tasks of Point clouds, Generating Point clouds remains challenging. This paper proposes a principled probabilistic framework to generate 3D Point clouds by modeling them as a distribution of distributions. Specifically, we learn a two-level hierarchy of distributions where the first level is the distribution of shapes and the second level is the distribution of Points given a shape. This formulation allows us to both sample shapes and sample an arbitrary number of Points from a shape. Our generative model, named PointFlow, learns each level of the distribution with a continuous normalizing flow. The invertibility of normalizing flows enables the computation of the likelihood during training and allows us to train our model in the variational inference framework. Empirically, we demonstrate that PointFlow achieves state-of-the-art performance in Point cloud generation. We additionally show that our model can faithfully reconstruct Point clouds and learn useful representations in an unsupervised manner. The code is available at https://github.com/stevenygd/PointFlow.
-
Pointflow 3d Point cloud generation with continuous normalizing flows
arXiv: Computer Vision and Pattern Recognition, 2019Co-Authors: Guandao Yang, Xun Huang, Serge Belongie, Bharath HariharanAbstract:As 3D Point clouds become the representation of choice for multiple vision and graphics applications, the ability to synthesize or reconstruct high-resolution, high-fidelity Point clouds becomes crucial. Despite the recent success of deep learning models in discriminative tasks of Point clouds, Generating Point clouds remains challenging. This paper proposes a principled probabilistic framework to generate 3D Point clouds by modeling them as a distribution of distributions. Specifically, we learn a two-level hierarchy of distributions where the first level is the distribution of shapes and the second level is the distribution of Points given a shape. This formulation allows us to both sample shapes and sample an arbitrary number of Points from a shape. Our generative model, named PointFlow, learns each level of the distribution with a continuous normalizing flow. The invertibility of normalizing flows enables the computation of the likelihood during training and allows us to train our model in the variational inference framework. Empirically, we demonstrate that PointFlow achieves state-of-the-art performance in Point cloud generation. We additionally show that our model can faithfully reconstruct Point clouds and learn useful representations in an unsupervised manner. The code will be available at this https URL.
Rui Wang - One of the best experts on this subject based on the ideXlab platform.
-
multiresolution tree networks for 3d Point cloud processing
arXiv: Computer Vision and Pattern Recognition, 2018Co-Authors: Matheus Gadelha, Rui Wang, Subhransu MajiAbstract:We present multiresolution tree-structured networks to process Point clouds for 3D shape understanding and generation tasks. Our network represents a 3D shape as a set of locality-preserving 1D ordered list of Points at multiple resolutions. This allows efficient feed-forward processing through 1D convolutions, coarse-to-fine analysis through a multi-grid architecture, and it leads to faster convergence and small memory footprint during training. The proposed tree-structured encoders can be used to classify shapes and outperform existing Point-based architectures on shape classification benchmarks, while tree-structured decoders can be used for Generating Point clouds directly and they outperform existing approaches for image-to-shape inference tasks learned using the ShapeNet dataset. Our model also allows unsupervised learning of Point-cloud based shapes by using a variational autoencoder, leading to higher-quality generated shapes.
-
Point sampling with general noise spectrum
International Conference on Computer Graphics and Interactive Techniques, 2012Co-Authors: Yahan Zhou, Haibin Huang, Liyi Wei, Rui WangAbstract:Point samples with different spectral noise properties (often defined using color names such as white, blue, green, and red) are important for many science and engineering disciplines including computer graphics. While existing techniques can easily produce white and blue noise samples, relatively little is known for Generating other noise patterns. In particular, no single algorithm is available to generate different noise patterns according to user-defined spectra. In this paper, we describe an algorithm for Generating Point samples that match a user-defined Fourier spectrum function. Such a spectrum function can be either obtained from a known sampling method, or completely constructed by the user. Our key idea is to convert the Fourier spectrum function into a differential distribution function that describes the samples' local spatial statistics; we then use a gradient descent solver to iteratively compute a sample set that matches the target differential distribution function. Our algorithm can be easily modified to achieve adaptive sampling, and we provide a GPU-based implementation. Finally, we present a variety of different sample patterns obtained using our algorithm, and demonstrate suitable applications.
Schoenberg Ronny - One of the best experts on this subject based on the ideXlab platform.
-
Pervasively anoxic surface conditions at the onset of the Great Oxidation Event: new multi-proxy constraints from the Cooper Lake paleosol
Iowa State University Digital Repository, 2019Co-Authors: Babechuk, Michael G., Weimar Nadine, Kleinhanns, Ilka C., Eroglu Suemeyya, Swanner, Elizabeth D., Kenny, Gavin G., Kamber, Balz S., Schoenberg RonnyAbstract:Oceanic element inventories derived from marine sedimentary rocks place important constraints on oxidative continental weathering in deep time, but there remains a scarcity in complementary observations directly from continental sedimentary reservoirs. This study focuses on better defining continental weathering conditions near the Archean-Proterozoic boundary through the multi-proxy (major and ultra-trace element, Fe and Cr stable isotopes, μ-XRF elemental mapping, and detrital zircon U-Pb geochronology) investigation of the ca. 2.45 billion year old (giga annum, Ga) Cooper Lake paleosol (saprolith), developed on a sediment-hosted mafic dike within the Huronian Supergroup (Ontario, Canada). Throughout the variably altered Cooper Lake saprolith, ratios of immobile elements (Nb, Ta, Zr, Hf, Th, Al, Ti) are constant, indicating a uniform pre-alteration dike composition, lack of extreme pH weathering conditions, and no major influence from ligand-rich fluids during weathering or burial metasomatism/metamorphism. The loss of Mg, Fe, Na, Sr, and Li, a signature of albite and ferromagnesian silicate weathering, increases towards the top of the preserved profile (unconformity) and dike margins. Coupled bulk rock behaviour of Fe-Mg-Mn and co-localization of Fe- Mn in clay minerals (predominantly chlorite) indicates these elements were solubilized primarily in their divalent state without Fe/Mn-oxide formation. A lack of a Ce anomaly and immobility of Mo, V, and Cr further support pervasively anoxic weathering conditions. Subtle U enrichment is the only geochemical evidence, if primary, that could be consistent with oxidative element mobilization. The leaching of ferromagnesian silicates was accompanied by variable mobility and depletion of transition metals with a relative depletion order of Fe≈Mg≈Zn\u3eNi\u3eCo\u3eCu (Cu being significantly influenced by secondary sulfide formation). Mild enrichment of heavy Fe isotopes (δ56/54Fe from 0.169 to 0.492 ‰) correlating with Fe depletion in the saprolith indicates loss of isotopically light aqueous Fe(II). Minor REE+Y fractionation with increasing alteration intensity, including a decreasing Eu anomaly and Y/Ho ratio, is attributed to albite breakdown and preferential scavenging of HREE\u3eY by clay minerals, respectively. Younger metasomatism resulted in the addition of several elements (K, Rb, Cs, Be, Tl, Ba, Sn, In, W), partly or wholly obscuring their earlier paleo-weathering trends. The behavior of Cr at Cooper Lake can help test previous hypotheses of an enhanced, low pH-driven continental weathering flux of Cr(III) to marine reservoirs between ca. 2.48-2.32 Ga and the utility of the stable Cr isotope proxy of Mn-oxide induced Cr(III) oxidation. Synchrotron μ- XRF maps and invariant Cr/Nb ratios reveal complete immobility of Cr despite its distribution amongst both clay-rich groundmass and Fe-Ti oxides. Assuming a pH-dependent, continental source of Cr(III) to marine basins, the Cr immobility at Cooper Lake indicates either that signatures of acidic surface waters were localized to uppermost and typically unpreserved regolith horizons or were geographically restricted to acid-Generating Point sources. However, in given detrital pyrite preservation in fluvial sequences overlying the paleosol, we propose that the oxidative sulphide corrosion required to drive surface pH(δ53/52Cr: -0.321 ± 0.038 ‰, 2sd, n=34) that cannot be linked to Cr(III) oxidation and is instead interpreted to have a magmatic origin. The combined chemical signatures and continued preservation of detrital pyrite/uraninite indicate low atmospheric O2 during weathering at ca. 2.45 Ga preserved in the rift-related sedimentary rocks of the Lower Huronian. The aqueous flux from the reduced weathering of mafic rocks was characterized by a greater abundance of transition metals (Fe, Mn, Zn, Co, Ni) with isotopically light Fe(II), as well as higher Eu/Eu* and Y/Ho. In most models of Precambrian ocean element inventories, hydrothermal fluids are viewed as the main supplier of several metals (e.g., Fe, Zn), although the results herein suggest that a riverine metal supply may have been substantial and that using Eu-excess as a strict proxy for hydrothermal flux may be misleading in near-shore marine sedimentary environments
-
Pervasively anoxic surface conditions at the onset of the Great Oxidation Event: New multi-proxy constraints from the Cooper Lake paleosol
'Elsevier BV', 2019Co-Authors: Babechuk, Michael G., Kleinhanns, Ilka C., Eroglu Suemeyya, Swanner, Elizabeth D., Kenny, Gavin G., Kamber, Balz S., Weimar, Nadine E., Schoenberg RonnyAbstract:Oceanic element inventories derived from marine sedimentary rocks place important constraints on oxidative continental weathering in deep time, but there remains a scarcity in complementary observations directly from continental sedimentary reservoirs. This study focuses on better defining continental weathering conditions near the Archean-Proterozoic boundary through the multi-proxy (major and ultra-trace element, Fe and Cr stable isotopes, µ-XRF elemental mapping, and detrital zircon U-Pb geochronology) investigation of the ca. 2.45 billion year old (giga annum, Ga) Cooper Lake paleosol (saprolith) developed on a sediment-hosted mafic dike within the Huronian Supergroup (Ontario, Canada). Throughout the variably altered Cooper Lake saprolith, ratios of immobile elements (Nb, Ta, Zr, Hf, Th, Al, Ti) are constant, indicating a uniform pre-alteration dike composition, lack of extreme pH weathering conditions, and no major influence from ligand-rich fluids during weathering or burial metasomatism/metamorphism. The loss of Mg, Fe, Na, Sr, and Li, a signature of albite and ferromagnesian silicate weathering, increases towards the top of the preserved profile (unconformity) and dike margins. Coupled bulk rock behaviour of Fe-Mg-Mn and co-localization of Fe-Mn in clay minerals (predominantly chlorite) indicates these elements were solubilized primarily in their divalent state without Fe/Mn-oxide formation. A lack of a Ce anomaly and immobility of Mo, V, and Cr further support pervasively anoxic weathering conditions. Subtle U enrichment, if primary, is the only geochemical evidence that could be consistent with oxidative element mobilization. The leaching of ferromagnesian silicates was accompanied by variable mobility and depletion of transition metals with a relative depletion order of Fe ≈ Mg ≈ Zn > Ni > Co > Cu (Cu being significantly influenced by secondary sulfide formation). Mild enrichment of heavy Fe isotopes (δ 56/54 Fe from 0.169 to 0.492‰) correlating with Fe depletion in the saprolith indicates open-system loss of isotopically light aqueous Fe(II). Minor REE + Y fractionation with increasing alteration intensity, including a decreasing Eu anomaly and Y/Ho ratio, is attributed to albite breakdown and preferential scavenging of HREE > Y by clay minerals, respectively. Younger metasomatism resulted in the addition of several elements (K, Rb, Cs, Be, Tl, Ba, Sn, In, W), partly or wholly obscuring their earlier paleo-weathering trends. The behavior of Cr at Cooper Lake can help test previous hypotheses of an enhanced, low pH-driven continental weathering flux of Cr(III) to marine reservoirs between ca. 2.48–2.32 Ga and the utility of the stable Cr isotope proxy of Mn-oxide induced Cr(III) oxidation. Synchrotron µ-XRF maps and invariant Cr/Nb ratios reveal complete immobility of Cr despite its distribution amongst both clay-rich groundmass and Fe-Ti oxides. Assuming a pH-dependent, continental source of Cr(III) to marine basins, the Cr immobility at Cooper Lake indicates either that signatures of acidic surface waters were localized to uppermost and typically unpreserved regolith horizons or were geographically restricted to acid-Generating Point sources. However, given detrital pyrite preservation in overlying fluvial sequences, it is probable that the oxidative sulfide corrosion required to drive surface pH < 4 lagged behind in this region relative to other early Proterozoic sequences. The entire saprolith exhibits a consistently light stable Cr isotope composition (δ 53/52 Cr: −0.321 ± 0.038‰ 2sd, n = 34) that cannot be linked to Cr(III) oxidation and is instead interpreted to have a magmatic origin. The combined paleosol chemical signatures and preservation of detrital pyrite/uraninite indicate low atmospheric O 2 during weathering at ca. 2.45 Ga in the rift-related environment of the Lower Huronian. The aqueous flux from the reduced weathering of mafic rocks was presumably characterized by a greater transition metal (Fe, Mn, Zn, Co, Ni) load with isotopically light Fe(II) compared to modern environments, as well as higher Eu/Eu * and Y/Ho than the source rock. In most models of Precambrian ocean element inventories, hydrothermal fluids are viewed as the main supplier of several metals (e.g., Fe, Zn), although the results herein suggest that a riverine metal supply may have been substantial and that using Eu-excess as a strict proxy for hydrothermal flux may be misleading in near-shore marine sedimentary deposits
Guandao Yang - One of the best experts on this subject based on the ideXlab platform.
-
Pointflow 3d Point cloud generation with continuous normalizing flows
International Conference on Computer Vision, 2019Co-Authors: Guandao Yang, Xun Huang, Serge Belongie, Bharath HariharanAbstract:As 3D Point clouds become the representation of choice for multiple vision and graphics applications, the ability to synthesize or reconstruct high-resolution, high-fidelity Point clouds becomes crucial. Despite the recent success of deep learning models in discriminative tasks of Point clouds, Generating Point clouds remains challenging. This paper proposes a principled probabilistic framework to generate 3D Point clouds by modeling them as a distribution of distributions. Specifically, we learn a two-level hierarchy of distributions where the first level is the distribution of shapes and the second level is the distribution of Points given a shape. This formulation allows us to both sample shapes and sample an arbitrary number of Points from a shape. Our generative model, named PointFlow, learns each level of the distribution with a continuous normalizing flow. The invertibility of normalizing flows enables the computation of the likelihood during training and allows us to train our model in the variational inference framework. Empirically, we demonstrate that PointFlow achieves state-of-the-art performance in Point cloud generation. We additionally show that our model can faithfully reconstruct Point clouds and learn useful representations in an unsupervised manner. The code is available at https://github.com/stevenygd/PointFlow.
-
Pointflow 3d Point cloud generation with continuous normalizing flows
arXiv: Computer Vision and Pattern Recognition, 2019Co-Authors: Guandao Yang, Xun Huang, Serge Belongie, Bharath HariharanAbstract:As 3D Point clouds become the representation of choice for multiple vision and graphics applications, the ability to synthesize or reconstruct high-resolution, high-fidelity Point clouds becomes crucial. Despite the recent success of deep learning models in discriminative tasks of Point clouds, Generating Point clouds remains challenging. This paper proposes a principled probabilistic framework to generate 3D Point clouds by modeling them as a distribution of distributions. Specifically, we learn a two-level hierarchy of distributions where the first level is the distribution of shapes and the second level is the distribution of Points given a shape. This formulation allows us to both sample shapes and sample an arbitrary number of Points from a shape. Our generative model, named PointFlow, learns each level of the distribution with a continuous normalizing flow. The invertibility of normalizing flows enables the computation of the likelihood during training and allows us to train our model in the variational inference framework. Empirically, we demonstrate that PointFlow achieves state-of-the-art performance in Point cloud generation. We additionally show that our model can faithfully reconstruct Point clouds and learn useful representations in an unsupervised manner. The code will be available at this https URL.
Raanan Fattal - One of the best experts on this subject based on the ideXlab platform.
-
blue noise Point sampling using kernel density model
International Conference on Computer Graphics and Interactive Techniques, 2011Co-Authors: Raanan FattalAbstract:Stochastic Point distributions with blue-noise spectrum are used extensively in computer graphics for various applications such as avoiding aliasing artifacts in ray tracing, halftoning, stippling, etc. In this paper we present a new approach for Generating Point sets with high-quality blue noise properties that formulates the problem using a statistical mechanics interacting particle model. Points distributions are generated by sampling this model. This new formulation of the problem unifies randomness with the requirement for equidistant Point spacing, responsible for the enhanced blue noise spectral properties. We derive a highly efficient multi-scale sampling scheme for drawing random Point distributions from this model. The new scheme avoids the critical slowing down phenomena that plagues this type of models. This derivation is accompanied by a model-specific analysis. Altogether, our approach generates high-quality Point distributions, supports spatially-varying spatial Point density, and runs in time that is linear in the number of Points generated.