91,250 Hits in 9.3 sec

Generalized Shape Metrics on Neural Representations [article]

Alex H. Williams and Erin Kunz and Simon Kornblith and Scott W. Linderman
2022 arXiv   pre-print
Using this framework we modify existing representational similarity measures based on canonical correlation analysis to satisfy the triangle inequality, formulate a novel metric that respects the inductive  ...  To identify general principles, researchers are increasingly interested in surveying large collections of networks that are trained on, or biologically adapted to, similar tasks.  ...  Supplemental Information: Generalized Shape Metrics on Neural Representations This supplement is organized into five sections.  ... 
arXiv:2110.14739v2 fatcat:apz7xacyjnb6vgy6no3dx5hfza

Abstract representations emerge naturally in neural networks trained to perform multiple tasks [article]

W. Jeffrey Johnston, Stefano Fusi
2021 bioRxiv   pre-print
We posit that this ability depends on the geometry of the neural population representations of these objects and contexts.  ...  Specifically, abstract, or disentangled, neural representations -- in which neural population activity is a linear function of the variables important for making a decision -- are known to allow for this  ...  Acknowledgments: We thank Mattia Rigotti, Nicolas Masse, and Matthew for comments on an earlier version of this manuscript.  ... 
doi:10.1101/2021.10.20.465187 fatcat:5wfchccfjzhwvmoxwhfv5bwt64

DeepCurrents: Learning Implicit Representations of Shapes with Boundaries [article]

David Palmer and Dmitriy Smirnov and Stephanie Wang and Albert Chern and Justin Solomon
2022 arXiv   pre-print
We propose a hybrid shape representation that combines explicit boundary curves with implicit learned interiors.  ...  By modifying the metric according to target geometry coming, e.g., from a mesh or point cloud, we can use this approach to represent arbitrary surfaces, learning implicitly defined shapes with explicitly  ...  For now, we will assume the metric is Euclidean; see Section 4.2 for the generalization to Riemannian metrics.  ... 
arXiv:2111.09383v2 fatcat:l7jv62o5jrevddktahtmph5n5e

On the Effectiveness of Weight-Encoded Neural Implicit 3D Shapes [article]

Thomas Davies and Derek Nowrouzezahrai and Alec Jacobson
2021 arXiv   pre-print
as a 3D shape representation.  ...  Many prior works have focused on _latent-encoded_ neural implicits, where a latent vector encoding of a specific shape is also fed as input.  ...  The signed distance function (SDF) of a surface can be defined by the metric set Ω of points within the shape, along with metric d.  ... 
arXiv:2009.09808v3 fatcat:7nvr4plxzbgevgqmtilk5zron4

Neural Template: Topology-aware Reconstruction and Disentangled Generation of 3D Meshes [article]

Ka-Hei Hui, Ruihui Li, Jingyu Hu, Chi-Wing Fu
2022 arXiv   pre-print
One key insight is to decouple the complex mesh reconstruction into two sub-tasks: topology formulation and shape deformation.  ...  Thanks to the decoupling, DT-Net implicitly learns a disentangled representation for the topology and shape in the latent space.  ...  In this work, we propose to implicitly learn a disentangled representation for the topology and shape, facilitating novel controls on the 3D mesh generation process.  ... 
arXiv:2206.04942v1 fatcat:ufrpzajkvza4ncci6csrfzqhca

DeformNet: Free-Form Deformation Network for 3D Shape Reconstruction from a Single Image [article]

Andrey Kurenkov, Jingwei Ji, Animesh Garg, Viraj Mehta, JunYoung Gwak, Christopher Choy, Silvio Savarese
2017 arXiv   pre-print
Prior methods have tackled this problem through generative models which predict 3D reconstructions as voxels or point clouds.  ...  DeformNet takes an image input, searches the nearest shape template from a database, and deforms the template to match the query image.  ...  We use a parametric metric space representation using a neural network F(·; θ ) where θ denote the parameters in the neural network.  ... 
arXiv:1708.04672v1 fatcat:tdbjw6rn5bch3kuidapivqgsd4

Neural Points: Point Cloud Representation with Neural Fields for Arbitrary Upsampling [article]

Wanquan Feng, Jin Li, Hongrui Cai, Xiaonan Luo, Juyong Zhang
2022 arXiv   pre-print
Therefore, Neural Points contain more shape information and thus have a stronger representation ability.  ...  Experimental results show that Neural Points has powerful representation ability and demonstrate excellent robustness and generalization ability.  ...  results demonstrate that the Neural Points representation has excellent robustness and generalization ability for various inputs on the upsampling task.  ... 
arXiv:2112.04148v3 fatcat:qm22oxc2inacpdslnzctxy2hri

NeuralODF: Learning Omnidirectional Distance Fields for 3D Shape Representation [article]

Trevor Houchens, Cheng-You Lu, Shivam Duggal, Rao Fu, Srinath Sridhar
2022 arXiv   pre-print
Experiments demonstrate that NeuralODF can learn to capture high-quality shape by overfitting to a single object, and also learn to generalize on common shape categories.  ...  Each representation is suited for different tasks thus making the transformation of one representation into another (forward map) an important and common problem.  ...  Neu-ralODF can generalize to unseen shapes with updated latent code. as a closed surface shape Iter. Depth Metrics Intersection Conf.  ... 
arXiv:2206.05837v3 fatcat:3eqfgmnuifczrayehrsqut7ft4

StableFace: Analyzing and Improving Motion Stability for Talking Face Generation [article]

Jun Ling, Xu Tan, Liyang Chen, Runnan Li, Yuchao Zhang, Sheng Zhao, Li Song
2022 arXiv   pre-print
; 2) we add augmented erosions on the input data of the neural renderer in training to simulate the distortion in inference to reduce mismatch; 3) we develop an audio-fused transformer generator to model  ...  In this paper, we conduct systematic analyses on the motion jittering problem based on a state-of-the-art pipeline that uses 3D face representations to bridge the input audio and output video, and improve  ...  First, jitters from the 3D face representations s 1:T . We adopt the choice of 3D face shapes to bridge the neural renderer and audio processing module.  ... 
arXiv:2208.13717v1 fatcat:6yvjayz5pjaatilpagxrdsdosm

Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes [article]

Towaki Takikawa, Joey Litalien, Kangxue Yin, Karsten Kreis, Charles Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, Sanja Fidler
2021 arXiv   pre-print
Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes.  ...  Furthermore, it produces state-of-the-art reconstruction quality for complex shapes under both 3D geometric and 2D image-space metrics.  ...  Generalization We now show that our surface extraction mechanism can generalize to multiple shapes, even from being trained on a single shape.  ... 
arXiv:2101.10994v1 fatcat:w3sgfbn4kzahtcr2uwnlaivgey

TriangleConv: A Deep Point Convolutional Network for Recognizing Building Shapes in Map Space

Chun Liu, Yaohui Hu, Zheng Li, Junkui Xu, Zhigang Han, Jianzhong Guo
2021 ISPRS International Journal of Geo-Information  
As buildings in map space are often represented as the vector data, research was conducted to learn the feature representations of the buildings and recognize their shapes based on graph neural networks  ...  The classification and recognition of the shapes of buildings in map space play an important role in spatial cognition, cartographic generalization, and map updating.  ...  Therefore, related work has been done to extract the deep feature representations of the buildings and recognize their shapes based on graph neural networks.  ... 
doi:10.3390/ijgi10100687 doaj:4e613379951349e9a5bca54999c7ab87 fatcat:mfws5djx4zczdhglek64gdjipi

Distinguishing Conjoint and Independent Neural Tuning for Stimulus Features With fMRI Adaptation

Daniel M. Drucker, Wesley Thomas Kerr, Geoffrey Karl Aguirre
2009 Journal of Neurophysiology  
evoked by changes in one versus two stimulus dimensions and considering the metric of two-dimension additivity.  ...  We describe an application of functional magnetic resonance imaging (fMRI) adaptation to distinguish between independent and conjoint neural representations of dimensions by examining the neural signal  ...  Extension to a generalized neural space metric So far we have considered two competing, extreme models of neural representation: two independent neural populations that represent stimulus dimensions separately  ... 
doi:10.1152/jn.91306.2008 pmid:19357342 pmcid:PMC2694123 fatcat:xpnpkul555fqrgfc2saasjvpya

Neural Lumigraph Rendering [article]

Petr Kellnhofer, Lars Jebe, Andrew Jones, Ryan Spicer, Kari Pulli, Gordon Wetzstein
2021 arXiv   pre-print
Our neural rendering pipeline accelerates SOTA neural volume rendering by about two orders of magnitude and our implicit surface representation is unique in allowing us to export a mesh with view-dependent  ...  We adopt high-capacity neural scene representations with periodic activations for jointly optimizing an implicit surface and a radiance field of a scene supervised exclusively with posed 2D images.  ...  Neural Volumes We use the original code shared by the authors [31] . We train our models for at least 150K iterations.  ... 
arXiv:2103.11571v1 fatcat:ltqjdj6it5brtgeilws2gipbnm

3DILG: Irregular Latent Grids for 3D Generative Modeling [article]

Biao Zhang, Matthias Nießner, Peter Wonka
2022 arXiv   pre-print
Existing works on neural fields are grid-based representations with latents defined on a regular grid.  ...  We propose a new representation for encoding 3D shapes as neural fields.  ...  Work Neural shape representations Shape analysis with neural networks processes shapes in different representations.  ... 
arXiv:2205.13914v1 fatcat:yykrn4xnwzbpnaqxaqfbqr73yi

Identifying Style of 3D Shapes using Deep Metric Learning

Isaak Lim, Anne Gehre, Leif Kobbelt
2016 Computer graphics forum (Print)  
We represent the shapes as rendered images and show how image tuples can be selected, generated and used efficiently for deep metric learning.  ...  This allows us to train the similarity metric on a shape collection directly, since any low-or high-level features needed to discriminate between different styles are identified by the neural network automatically  ...  Firstly, we have to choose a representation of 3D shapes on which to train our neural network.  ... 
doi:10.1111/cgf.12977 fatcat:uqid37fj5bel5k4rxsg3t6b3fi
« Previous Showing results 1 — 15 out of 91,250 results