Filters








544 Hits in 4.3 sec

Weakly Supervised High-Fidelity Clothing Model Generation [article]

Ruili Feng, Cheng Ma, Chengji Shen, Xin Gao, Zhenjiang Liu, Xiaobo Li, Kairi Ou, Zhengjun Zha
2021 arXiv   pre-print
Experiments on real scene proprietary model images demonstrate the superiority of DGP over several state-of-the-art supervised methods when generating clothing model images.  ...  In this paper, we propose a cheap yet scalable weakly-supervised method called Deep Generative Projection (DGP) to address this specific scenario.  ...  The proposed weakly-supervised method outperforms all supervised competitors significantly over all three aspects.  ... 
arXiv:2112.07200v1 fatcat:e2y2xrzw6fgv5n3maogrnwrjha

UVIRT—Unsupervised Virtual Try-on Using Disentangled Clothing and Person Features

Hideki Tsunashima, Kosuke Arase, Antony Lam, Hirokatsu Kataoka
2020 Sensors  
Even existing weakly-supervised virtual try-on methods still use annotated data or pre-trained networks as auxiliary information and the costs of the annotation are still significantly high.  ...  These approaches incur a very high cost in annotation.  ...  These adversarial losses encourage generating high fidelity images.  ... 
doi:10.3390/s20195647 pmid:33023177 pmcid:PMC7582289 fatcat:dsv442mrmnh7tefojzjs5risue

SCANimate: Weakly Supervised Learning of Skinned Clothed Avatar Networks [article]

Shunsuke Saito, Jinlong Yang, Qianli Ma, Michael J. Black
2021 arXiv   pre-print
We demonstrate our approach on various clothing types with different amounts of training data, outperforming existing solutions and other variants in terms of fidelity and generality in every setting.  ...  These observations lead us to a weakly supervised learning method that aligns scans into a canonical pose by disentangling articulated deformations without template-based surface registration.  ...  SCANimate learns a pose-aware parametric clothed human model directly from raw scans in a weakly supervised manner.  ... 
arXiv:2104.03313v2 fatcat:bz76iumdbzd6fn7v74tzuatfiy

Deep Physics-aware Inference of Cloth Deformation for Monocular Human Performance Capture [article]

Yue Li, Marc Habermann, Bernhard Thomaszewski, Stelian Coros, Thabo Beeler, Christian Theobalt
2021 arXiv   pre-print
of weakly supervised deep monocular human performance capture.  ...  We show how integrating physics into the training process improves the learned cloth deformations, allows modeling clothing as a separate piece of geometry, and largely reduces cloth-body intersections  ...  By using simulation supervision and separate modeling of cloth and body geometries, the wrinkles generated by our method are less constrained by the template and, consequently, exhibit more variety and  ... 
arXiv:2011.12866v2 fatcat:unwkgbplznatfjr2t66zt644dq

Deep Physics-aware Inference of Cloth Deformation for Monocular Human Performance Capture

Yue Li, Marc Habermann, Bernhard Thomaszewski, Stelian Coros, Thabo Beeler, Christian Theobalt
2021 2021 International Conference on 3D Vision (3DV)  
in the context of weakly-supervised deep monocular human performance capture.  ...  We show how integrating physics into the training process improves the learned cloth deformations, allows modeling clothing as a separate piece of geometry, and largely reduces cloth-body intersections  ...  By using simulation supervision and separate modeling of cloth and body geometries, the wrinkles generated by our method are less constrained by the template and, consequently, exhibit more variety and  ... 
doi:10.1109/3dv53792.2021.00047 fatcat:ddbya2wsczfntmt3hi45dnvrqa

Pose Guided Attention for Multi-label Fashion Image Classification [article]

Beatriz Quintino Ferreira, João P. Costeira, Ricardo G. Sousa, Liang-Yan Gui, João P. Gomes
2019 arXiv   pre-print
Our visual semantic attention model (VSAM) is supervised by automatic pose extraction creating a discriminative feature space.  ...  In our case, this mechanism acts on a feature combination scheme used in [12] and is supervised by the pose of the human model wearing the clothing item.  ...  However, the performance gains on the computing side usually come with a high cost from human-intensive tasks to generate high quality training data, a key (and expensive) issue in the fashion industry  ... 
arXiv:1911.05024v1 fatcat:cucamml42vg6vpbe3zj4t7nguu

Multi-View Consistency Loss for Improved Single-Image 3D Reconstruction of Clothed People [article]

Akin Caliskan, Armin Mustafa, Evren Imre, Adrian Hilton
2020 arXiv   pre-print
The 3DVH dataset of realistic clothed 3D human models rendered with diverse natural backgrounds is demonstrated to allows transfer to reconstruction from real images of people.  ...  Recent work has introduced volumetric, implicit and model-based shape learning frameworks for reconstruction of objects and people from one or more images.  ...  In order to estimate high-fidelity and accurate 3D models on real images, the synthetic rendering should be as realistic as possible [13] .  ... 
arXiv:2009.14162v1 fatcat:aqrjlw4djvgdrpcxdcwwcsaaee

2021 Index IEEE Transactions on Pattern Analysis and Machine Intelligence Vol. 43

2022 IEEE Transactions on Pattern Analysis and Machine Intelligence  
Zhou, P., +, TPAMI May 2021 1718-1732 Weakly-Supervised Learning of Category-Specific 3D Object Shapes.  ...  Chen, R., +, TPAMI Oct. 2021 3695-3708 Weakly-Supervised Learning of Category-Specific 3D Object Shapes.  ...  Grammars A Generalized Earley Parser for Human Activity Parsing and Prediction. Qi, S., +, TPAMI Aug. 2021 Damen, D., +, TPAMI Nov. 2021 4125-4141  ... 
doi:10.1109/tpami.2021.3126216 fatcat:h6bdbf2tdngefjgj76cudpoyia

A Deeper Look into DeepCap [article]

Marc Habermann, Weipeng Xu, Michael Zollhoefer, Gerard Pons-Moll, Christian Theobalt
2021 arXiv   pre-print
Our method is trained in a weakly supervised manner based on multi-view supervision completely removing the need for training data with 3D ground truth annotations.  ...  To avoid generating dense 3D ground truth annotation, our network is trained in a weakly supervised manner.  ...  Since generating the ground truth for θ and α is a nontrivial task, we propose weakly supervised training based on fitting the skeleton to multi-view 2D joint detections. Kinematics Layer.  ... 
arXiv:2111.10563v1 fatcat:pkwf5736rje43ihu7pmkjjggly

Image Synthesis From Reconfigurable Layout and Style [article]

Wei Sun, Tianfu Wu
2019 arXiv   pre-print
Inspired by the vanilla StyleGAN, the proposed LostGAN consists of two new components: (i) learning fine-grained mask maps in a weakly-supervised manner to bridge the gap between layouts and images, and  ...  Despite remarkable recent progress on both unconditional and conditional image synthesis, it remains a long-standing problem to learn generative models that are capable of synthesizing realistic and sharp  ...  Unlike [12, 21] where ground truth masks is adopted to guide learning of shape generator, our model can learn semantic masks in a weakly-supervised manner.  ... 
arXiv:1908.07500v1 fatcat:hzvw33cy4vck5ahwnvmjg6e6ru

Adversarial Label Learning [article]

Chidubem Arachie, Bert Huang
2019 arXiv   pre-print
Experiments on three real datasets show that our method can train without labels and outperforms other approaches for weakly supervised learning.  ...  We propose a weakly supervised method---adversarial label learning---that trains classifiers to perform well against an adversary that chooses labels for training data.  ...  Our work is related to existing methods that use variants of a generalized expectation (GE) criteria (Druck, Mann, and McCallum 2008; for semi-and weakly supervised learning.  ... 
arXiv:1805.08877v3 fatcat:hhvpfxn4zna55if6d3txezswbu

A General Framework for Adversarial Label Learning

Chidubem Arachie, Bert Huang
2021 Journal of machine learning research  
Our experiments show that our method can train without labels and outperforms other approaches for weakly supervised learning.  ...  We propose a weakly supervised method-adversarial label learning-that trains classifiers to perform well when noisy and possibly correlated labels are provided.  ...  However, some existing approaches for weakly supervised learning (Ratner et al., 2017 (Ratner et al., , 2018;; Halpern et al., 2016) use generative modeling to model the possible correlations of the  ... 
dblp:journals/jmlr/ArachieH21 fatcat:ob46gpzd55boreptmr7erlabbu

Survey on depth and RGB image-based 3D hand shape and pose estimation

Lin Huang, Boshen Zhang, Zhilin Guo, Yang Xiao, Zhiguo Cao, Junsong Yuan
2021 Virtual Reality & Intelligent Hardware  
Parallel to 3D model renderings, model-based generative methods employ adversarial networks to generate high-fidelity synthetic depth images.  ...  Discriminative approaches: Weakly supervised, self-supervised, and semi-supervised learning Another way to alleviate the data-hungry problem is to conduct weakly supervised or self-supervised learning.  ... 
doi:10.1016/j.vrih.2021.05.002 fatcat:4tbhftt3ira6fporaqlscqhsse

Adversarial Label Learning

Chidubem Arachie, Bert Huang
2019 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Experiments on real datasets show that our method can train without labels and outperforms other approaches for weakly supervised learning.  ...  We propose a weakly supervised method—adversarial label learning—that trains classifiers to perform well against an adversary that chooses labels for training data.  ...  Training on more features could increase the predictive accuracy of the weak signals and by extension ALL, but such high-fidelity weak signals may be rare in practice.  ... 
doi:10.1609/aaai.v33i01.33013183 fatcat:xgymvk4ekzaitgw6eptr35g3va

DenseRaC: Joint 3D Pose and Shape Estimation by Dense Render-and-Compare

Yuanlu Xu, Song-Chun Zhu, Tony Tung
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
The generated data covers diversified camera views, human actions and body shapes, and is paired with full ground truth.  ...  Our model jointly learns to represent the 3D human body from hybrid datasets, mitigating the problem of unpaired training data.  ...  We would like to thank Tengyu Liu and Elan Markowitz for helping with data collection, Tuur Jan M Stuyck and Aaron Ferguson for cloth simulation, Natalia Neverova and colleagues at FRL, FAIR and UCLA for  ... 
doi:10.1109/iccv.2019.00785 dblp:conf/iccv/XuZT19 fatcat:u3c3lz62ezcyhev3fcbooo6lem
« Previous Showing results 1 — 15 out of 544 results