1,242 Hits in 5.0 sec

Unsupervised 3D Shape Completion through GAN Inversion [article]

Junzhe Zhang, Xinyi Chen, Zhongang Cai, Liang Pan, Haiyu Zhao, Shuai Yi, Chai Kiat Yeo, Bo Dai, Chen Change Loy
2021 arXiv   pre-print
Most 3D shape completion approaches rely heavily on partial-complete shape pairs and learn in a fully supervised manner.  ...  In contrast to previous fully supervised approaches, in this paper we present ShapeInversion, which introduces Generative Adversarial Network (GAN) inversion to shape completion for the first time.  ...  The GAN prior can be exploited through GAN inversion. Despite its notable success in various image restoration and manipulation tasks, it has not been explored for shape completion.  ... 
arXiv:2104.13366v2 fatcat:ysveo76go5b4dmqoitt54ndh6a

Learning a Structured Latent Space for Unsupervised Point Cloud Completion [article]

Yingjie Cai, Kwan-Yee Lin, Chao Zhang, Qiang Wang, Xiaogang Wang, Hongsheng Li
2022 arXiv   pre-print
Unsupervised point cloud completion aims at estimating the corresponding complete point cloud of a partial point cloud in an unpaired manner.  ...  By establishing such a unified and structured latent space, better partial-complete geometry consistency and shape completion accuracy can be achieved.  ...  A representative work [52] adopts GAN inversion for 3D shape completion. It trains a complete point cloud generator with adversarial losses.  ... 
arXiv:2203.15580v1 fatcat:kknlo6nfmjhifhigyxnptrbp5m

Spectral-GANs for High-Resolution 3D Point-cloud Generation [article]

Sameera Ramasinghe, Salman Khan, Nick Barnes, Stephen Gould
2020 arXiv   pre-print
Additionally, it can learn a highly discriminative representation in an unsupervised fashion and can be used to accurately reconstruct 3D objects.  ...  In this paper, we develop a principled approach to synthesize 3D point-clouds using a spectral-domain Generative Adversarial Network (GAN).  ...  79.9% Vconv-DAE (ECCV'16) [20] Unsupervised 80.5% 3D-GAN (NIPS'16) [28] Unsupervised 91.0% 3D-DesNet (CVPR'18) [31] Unsupervised 92.4% 3D-WINN (AAAI '19) [9] Unsupervised 91.9% PrimtiveGAN  ... 
arXiv:1912.01800v2 fatcat:hu6u7jhq3fgxvattdooez6gbyq

Inverse Graphics: Unsupervised Learning of 3D Shapes from Single Images [article]

Talip Ucar
2019 arXiv   pre-print
Forth, we present a general end-to-end approach to learning 3D shapes from single images in a completely unsupervised fashion by modelling the factors of variation such as azimuth as independent latent  ...  In this paper, we study the problem of unsupervised learning of 3D geometry from single images.  ...  Discussion In this work, we investigated unsupervised learning of 3D shapes from single images, using diverse set of datasets.  ... 
arXiv:1911.07937v2 fatcat:tr4zv3tipfe45kijtvys5nvtn4

StylePart: Image-based Shape Part Manipulation [article]

I-Chao Shen, Li-Wen Su, Yu-Ting Wu, Bing-Yu Chen
2022 arXiv   pre-print
Our method "forwardly maps" the image content to its corresponding 3D shape attributes, where the shape part can be easily manipulated.  ...  Our key contribution is a shape-consistent latent mapping function that connects the image generative latent space and the 3D man-made shape attribute latent space.  ...  Moreover, the shape parts are more complex than simple cylinders and cuboids. GAN inversion and latent space manipulation GAN inversion is required to edit a real image through latent manipulation.  ... 
arXiv:2111.10520v3 fatcat:u4rdoz6osfhmvfpdmu4iuhecem

A Survey on Face Data Augmentation [article]

Xiang Wang and Kai Wang and Shiguo Lian
2019 arXiv   pre-print
The Info-GAN proposed in [31] learned disentangled representations of faces in a completely unsupervised manner, and was able to modify the presence of glasses. Shen et al.  ...  [94] classified the 3d face models into 3DSM (3D Shape Model), 3DMM and E-3DMM (Extended 3DMM). The 3DSM can only explicitly model pose.  ... 
arXiv:1904.11685v1 fatcat:phcwwc7gcfablgytt6itr6xade

A Novel Technique to Regenerate Sculpture Using Generative Adversarial Network [chapter]

S.A.K. Jainulabudeen, H. Shalma, S. Gowri Shankar, D. Anuradha, K. Soniya
2020 Advances in Parallel Computing  
The art lovers of the present age can seek the knowledge of lost art through this modern day technology.  ...  So at present we do not allow the art to continue to fall into shadow and extinguish later on, thus in this paper we present a DC-GAN model which has been created to inherit all the artistic skills of  ...  This project can be enhanced by convolutional 3D and a 3D model of the sculpture can be generated and can be 3D printed. Figure 1 . 1 Figure 1. GAN Architecture Figure 2.  ... 
doi:10.3233/apc200149 fatcat:oqno3yak3nbwvpiv4goyrozkne

pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesis [article]

Eric R. Chan, Marco Monteiro, Petr Kellnhofer, Jiajun Wu, Gordon Wetzstein
2021 arXiv   pre-print
We propose a novel generative model, named Periodic Implicit Generative Adversarial Networks (π-GAN or pi-GAN), for high-quality 3D-aware image synthesis. π-GAN leverages neural representations with periodic  ...  activation functions and volumetric rendering to represent scenes as view-consistent 3D representations with fine detail.  ...  Although the unsupervised learning of 3D shapes was not the focus of this work, π-GAN nevertheless produces interpretable and view-consistent 3D representations that capture the 3D structures of objects  ... 
arXiv:2012.00926v2 fatcat:scgeqv3a4ngh7n4qzu2tkst5aq

Comprehensive Review of Deep Learning-Based 3D Point Cloud Completion Processing and Analysis [article]

Ben Fei, Weidong Yang, Wenming Chen, Zhijun Li, Yikang Li, Tao Ma, Xing Hu, Lipeng Ma
2022 arXiv   pre-print
Point cloud completion is a generation and estimation issue derived from the partial point clouds, which plays a vital role in the applications in 3D computer vision.  ...  The progress of deep learning (DL) has impressively improved the capability and robustness of point cloud completion.  ...  [87] presented ShapeInversion to introduce GAN inversion into shape completion for the first time.  ... 
arXiv:2203.03311v2 fatcat:e2kvryolufearetp4ujlw2gwwy

InGAN: Capturing and Remapping the "DNA" of a Natural Image [article]

Assaf Shocher, Shai Bagon, Phillip Isola, Michal Irani
2019 arXiv   pre-print
InGAN is fully unsupervised, requiring no additional data other than the input image itself.  ...  In this paper we propose an "Internal GAN" (InGAN) - an image-specific GAN - which trains on a single input image and learns its internal distribution of patches.  ...  without any 3D estimation).  ... 
arXiv:1812.00231v2 fatcat:sfiusmquzjd3rgs45lryz4l24u

InGAN: Capturing and Retargeting the "DNA" of a Natural Image

Assaf Shocher, Shai Bagon, Phillip Isola, Michal Irani
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
InGAN is fully unsupervised, requiring no additional data other than the input image itself.  ...  In this paper we propose an "Internal GAN" (InGAN) -an image-specific GAN -which trains on a single input image and learns its internal distribution of patches.  ...  without any 3D estimation).  ... 
doi:10.1109/iccv.2019.00459 dblp:conf/iccv/ShocherBII19 fatcat:jlj5vsmttzgkjabzsqjclsou2u

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
This formulation allows us to both sample shapes and sample an arbitrary number of points from a shape.  ...  We additionally show that our model can faithfully reconstruct point clouds and learn useful representations in an unsupervised manner.  ...  We then compute the probability of z in the prior distribution (L prior ) through a inverse CNF F −1 ψ , and compute the reconstruction likelihood of X (L recon ) through another inverse CNF G −1 θ conditioned  ... 
arXiv:1906.12320v3 fatcat:u2ilsq2fsfb77m653oeatt4674

Closed-Form Factorization of Latent Semantics in GANs [article]

Yujun Shen, Bolei Zhou
2021 arXiv   pre-print
In this work, we examine the internal representation learned by GANs to reveal the underlying variation factors in an unsupervised manner.  ...  A rich set of interpretable dimensions has been shown to emerge in the latent space of the Generative Adversarial Networks (GANs) trained for synthesizing images.  ...  Our algorithm can be performed in a completely unsupervised fashion by efficiently investigating the weights of a pre-trained GAN generator.  ... 
arXiv:2007.06600v4 fatcat:y2jtbcp345gsvad6dec7lbv6ou

Unsupervised Depth Estimation, 3D Face Rotation and Replacement [article]

Joel Ruben Antony Moniz, Christopher Beckham, Simon Rajotte, Sina Honari, Christopher Pal
2018 arXiv   pre-print
We present an unsupervised approach for learning to estimate three dimensional (3D) facial structure from a single image while also predicting 3D viewpoint transformations that match a desired pose and  ...  Lastly, we identify certain shortcomings with our formulation, and explore adversarial image translation techniques as a post-processing step to re-synthesize complete head shots for faces re-targeted  ...  Hassner et. al [8] explore the use of a single unmodified 3D surface as an approximation to the shape of all input faces.  ... 
arXiv:1803.09202v5 fatcat:mxwesnohgjcufo4ycyx3xjumgy

An Effective Loss Function for Generating 3D Models from Single 2D Image without Rendering [article]

Nikola Zubić, Pietro Liò
2021 arXiv   pre-print
Finally, we perform a GAN-based texture mapping on a particular 3D mesh and produce a textured 3D mesh from a single 2D image.  ...  Current renderers use losses based on pixels between a rendered image of some 3D reconstructed object and ground-truth images from given matched viewpoints to optimise parameters of the 3D shape.  ...  obtained through complete supervision.  ... 
arXiv:2103.03390v2 fatcat:orkruz4lprhyfb34p22h6jsaqq
« Previous Showing results 1 — 15 out of 1,242 results