1,707 Hits in 7.5 sec

3D reconstruction from structured-light profilometry with dual-path hybrid network

Lei Wang, Dunqiang Lu, Ruowen Qiu, Jiaqing Tao
2022 EURASIP Journal on Advances in Signal Processing  
The proposed dual-path hybrid network provides an effective solution for structured-light 3D reconstruction and its practice in engineering.  ...  Compared with classical methods such as Fourier transform profilometry, many deep neural networks are utilized to restore 3D shape from single-shot structured light.  ...  By adding a dilation rate, dilated convolution inserts blanks between the elements of the convolution kernel, which expands the kernel for a larger receptive field.  ... 
doi:10.1186/s13634-022-00848-5 fatcat:56qiwz2xhbfqpirrsn5qlqfa4u

Convolutional neural networks: an overview and application in radiology

Rikiya Yamashita, Mizuho Nishio, Richard Kinh Gian Do, Kaori Togashi
2018 Insights into Imaging  
learn spatial hierarchies of features through a backpropagation algorithm. • Familiarity with the concepts and advantages, as well as limitations, of convolutional neural network is essential to leverage  ...  radiology. • Convolutional neural network is composed of multiple building blocks, such as convolution layers, pooling layers, and fully connected layers, and is designed to automatically and adaptively  ...  Another study [9] utilized two types of 3D U-net for segmenting liver and liver mass on 3D CT images, which was named cascaded fully convolutional neural networks; one type of U-net was used for segmentation  ... 
doi:10.1007/s13244-018-0639-9 pmid:29934920 fatcat:vbo6znqwjbax7h425choj2ikwm

iUNets: Fully invertible U-Nets with Learnable Up- and Downsampling [article]

Christian Etmann and Rihuan Ke and Carola-Bibiane Schönlieb
2020 arXiv   pre-print
For large-scale data, as it for example appears in 3D medical imaging, the U-Net however has prohibitive memory requirements.  ...  U-Nets have been established as a standard architecture for image-to-image learning problems such as segmentation and inverse problems in imaging.  ...  Acknowledgements The authors thank Sil van de Leemput for his help in using and extending MemCNN, as well as Jens Behrmann for useful discussions around normalizing flows.  ... 
arXiv:2005.05220v3 fatcat:fbc3pizba5h5vfwkjhlat4fuxq

3D Shapes Local Geometry Codes Learning with SDF [article]

Shun Yao, Fei Yang, Yongmei Cheng, Mikhail G. Mozerov
2021 arXiv   pre-print
A signed distance function (SDF) as the 3D shape description is one of the most effective approaches to represent 3D geometry for rendering and reconstruction.  ...  Our work is inspired by the state-of-the-art method DeepSDF that learns and analyzes the 3D shape as the iso-surface of its shell and this method has shown promising results especially in the 3D shape  ...  as a neural network (usually as multilayer perceptron network) with learnable parameters θ.  ... 
arXiv:2108.08593v1 fatcat:zcz24elyxrcdpikhvp4qr2tq2u

xQSM: Quantitative Susceptibility Mapping with Octave Convolutional and Noise Regularized Neural Networks [article]

Yang Gao, Xuanyu Zhu, Bradford A. Moffat, Rebecca Glarin, Alan H. Wilman, G. Bruce Pike, Stuart Crozier, Feng Liu, Hongfu Sun
2020 arXiv   pre-print
In this study, a new deep learning method for QSM reconstruction, namely xQSM, was designed by introducing modified state-of-the-art octave convolutional layers into the U-net backbone.  ...  The results from a numerical phantom, a simulated human brain, four in vivo healthy human subjects, a multiple sclerosis patient, a glioblastoma patient, as well as a healthy mouse brain showed that the  ...  Acknowledgment We thank Steffen Bollmann and Markus Barth for sharing their codes for generating the synthetic training dataset in the present work.  ... 
arXiv:2004.06281v3 fatcat:57tavx3wwfeh7nod3nb4gmktfu

End-to-End View Synthesis for Light Field Imaging with Pseudo 4DCNN [chapter]

Yunlong Wang, Fei Liu, Zilei Wang, Guangqi Hou, Zhenan Sun, Tieniu Tan
2018 Lecture Notes in Computer Science  
The key advantage is to efficiently synthesize dense 4D light fields from a sparse set of input views.  ...  Specifically, 2D strided convolutions operated on stacked EPIs and detail-restoration 3D CNNs connected with angular conversion are assembled to build the Pseudo 4DCNN.  ...  and K r is the learnable kernel.  ... 
doi:10.1007/978-3-030-01216-8_21 fatcat:3ztzunoo2zavfik2wdx5sad2u4

DSTnet: Deformable Spatio-Temporal Convolutional Residual Network for Video Super-Resolution

Anusha Khan, Allah Bux Sargano, Zulfiqar Habib
2021 Mathematics  
The proposed framework consists of 3D convolutional residual blocks decomposed into spatial and temporal (2+1) D streams.  ...  The key challenge for VSR lies in the effective exploitation of intra-frame spatial relation and temporal dependency between consecutive frames.  ...  This ensures that (2+1)D block learnable param-eters are not more than that required for a 3D convolution kernel.  ... 
doi:10.3390/math9222873 fatcat:dlh2t4fs25bbvp55wb3mbgt42y

Deferred Neural Rendering: Image Synthesis using Neural Textures [article]

Justus Thies and Michael Zollhöfer and Matthias Nießner
2019 arXiv   pre-print
For instance, we can synthesize temporally-consistent video re-renderings of recorded 3D scenes as our representation is inherently embedded in 3D space.  ...  To address this challenging problem, we introduce Deferred Neural Rendering, a new paradigm for image synthesis that combines the traditional graphics pipeline with learnable components.  ...  as in the respective encoder layer, kernel size is 4 and stride is 2.  ... 
arXiv:1904.12356v1 fatcat:r3vfiidi4fdgpnbr6phlvhsuuu

Three-dimensional Measurement Using Structured Light Based on Deep Learning

Tao Zhang, Jin Xing Niu, Shuo Liu, Tao Tao Pan, Brij B Gupta
2021 Computer systems science and engineering  
The weighting coefficient loss function is introduced to the multi-convolution neural network, and the point-cloud data are continuously optimized to obtain the 3D reconstruction model.  ...  Three-dimensional (3D) reconstruction using structured light projection has the characteristics of non-contact, high precision, easy operation, and strong real-time performance.  ...  Acknowledgement: We thank LetPub ( for its linguistic assistance during the preparation of this manuscript.  ... 
doi:10.32604/csse.2021.014181 fatcat:ielt6zdffvdqjoiyuqtnqb3hha

Learning Local Neighboring Structure for Robust 3D Shape Representation [article]

Zhongpai Gao, Junchi Yan, Guangtao Zhai, Juyong Zhang, Yiyan Yang, Xiaokang Yang
2020 arXiv   pre-print
The recent success of convolutional neural networks (CNNs) for structured data (e.g., images) suggests the value of adapting insight from CNN for 3D shapes.  ...  Various graph neural networks for 3D shapes have been developed with isotropic filters or predefined local coordinate systems to overcome the node inconsistency on graphs.  ...  The node attributes can also include additional coordinates such as color and vertex normal. In a deep neural network, the output of each layer is as the input for the subsequent layer.  ... 
arXiv:2004.09995v3 fatcat:m6lig6dp4rhyznsytxnacoghfe

Unrolled Primal-Dual Networks for Lensless Cameras [article]

Oliver Kingshott, Nick Antipa, Emrah Bostan, Kaan Akşit
2022 arXiv   pre-print
These image reconstruction models fall short in simulating lensless cameras truthfully as these models are not sophisticated enough to account for optical aberrations or scenes with depth variations.  ...  This improvement stems from our primary finding that embedding learnable forward and adjoint models in a learned primal-dual optimization framework can even improve the quality of reconstructed images  ...  phases of the project; Koray Kavaklı for his support in hardware prototype related figure and camera homography related software; Tim Weyrich for dedicating GPU resource.  ... 
arXiv:2203.04353v1 fatcat:hxuxy5p3mncsxhg4vbao7frsvq

NeuVV: Neural Volumetric Videos with Immersive Rendering and Editing [article]

Jiakai Zhang, Liao Wang, Xinhang Liu, Fuqiang Zhao, Minzhang Li, Haizhao Dai, Boyuan Zhang, Wei Yang, Lan Xu, Jingyi Yu
2022 arXiv   pre-print
The core of NeuVV is to efficiently encode a dynamic neural radiance field (NeRF) into renderable and editable primitives.  ...  We introduce two types of factorization schemes: a hyper-spherical harmonics (HH) decomposition for modeling smooth color variations over space and time and a learnable basis representation for modeling  ...  [Bozic et al. 2020 ] applies data-driven approaches for non-rigid 3D reconstruction.  ... 
arXiv:2202.06088v1 fatcat:23qn5ffx6raglmp363hz5iizne

Geometrically Principled Connections in Graph Neural Networks [article]

Shunwang Gong, Mehdi Bahri, Michael M. Bronstein, Stefanos Zafeiriou
2020 arXiv   pre-print
We hope our simple and effective approach will serve as a solid baseline and help ease future research in graph neural networks.  ...  In this paper, we argue geometry should remain the primary driving force behind innovation in the emerging field of geometric deep learning.  ...  Additionally, stacking convolutional layers is known to increase the receptive field, including in graph neural networks [51] .  ... 
arXiv:2004.02658v1 fatcat:5sp67hopy5aldb54lfndfx6pqy

Instant tissue field and magnetic susceptibility mapping from MR raw phase using Laplacian enabled deep neural networks [article]

Yang Gao, Zhuang Xiong, Amir Fazlollahi, Peter J Nestor, Viktor Vegh, Fatima Nasrallah, Craig Winter, G. Bruce Pike, Stuart Crozier, Feng Liu, Hongfu Sun
2022 arXiv   pre-print
This study develops a large-stencil Laplacian preprocessed deep learning-based neural network for near instant quantitative field and susceptibility mapping (i.e., iQFM and iQSM) from raw MR phase data  ...  However, the traditional QSM reconstruction pipeline involves multiple non-trivial steps, including phase unwrapping, background field removal, and dipole inversion.  ...  In this work, the discrete Laplace kernel in the Lap-Layer was a 27-point stencil discrete Laplace estimate (kernel size: 3×3×3), fixed during training; however, more learnable convolutional kernels could  ... 
arXiv:2111.07665v3 fatcat:ral2hjz2qfhxlkgceoe5qpe7lq

Neural 3D Morphable Models: Spiral Convolutional Networks for 3D Shape Representation Learning and Generation

Giorgos Bouritsas, Sergiy Bokhnyak, Stylianos Ploumpis, Stefanos Zafeiriou, Michael Bronstein
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
, such as 3D voxel grids or 2D views.  ...  Generative models for 3D geometric data arise in many important applications in 3D computer vision and graphics.  ...  We use spiral convolution as a basic building block for hierarchical intrinsic mesh autoencoders, which we coin Neural 3D Morphable Models.  ... 
doi:10.1109/iccv.2019.00731 dblp:conf/iccv/BouritsasBPZB19 fatcat:tjlcs7zthveltj5dww54yz6u6a
« Previous Showing results 1 — 15 out of 1,707 results