Filters








5,731 Hits in 3.2 sec

Deep Kinematic Pose Regression [article]

Xingyi Zhou, Xiao Sun, Wei Zhang, Shuang Liang, Yichen Wei
2016 arXiv   pre-print
In this work, we propose to directly embed a kinematic object model into the deep neutral network learning for general articulated object pose estimation.  ...  Learning articulated object pose is inherently difficult because the pose is high dimensional but has many structural constraints.  ...  Deep Kinematic Pose Estimation Kinematic Model An articulated object is modeled as a kinematic model. A kinematic model is composed of several bones and joints.  ... 
arXiv:1609.05317v1 fatcat:h2s6f2afsfa6lagtpcfmjn3svy

Learning Transferable Kinematic Dictionary for 3D Human Pose and Shape Reconstruction [article]

Ze Ma, Yifan Yao, Pan Ji, Chao Ma
2021 arXiv   pre-print
Estimating 3D human pose and shape from a single image is highly under-constrained.  ...  To address this ambiguity, we propose a novel prior, namely kinematic dictionary, which explicitly regularizes the solution space of relative 3D rotations of human joints in the kinematic tree.  ...  With the emergence of deep learning, considerable efforts have been made to deal with 3D pose estimation by learning deep regression networks (Chen and Ramanan 2017; Martinez et al. 2017; Moreno-Noguer  ... 
arXiv:2104.00953v2 fatcat:xd6tytqdlvb4hhbbddqc3s3h3a

Learning to Search on Manifolds for 3D Pose Estimation of Articulated Objects [article]

Yu Zhang, Chi Xu, Li Cheng
2016 arXiv   pre-print
This paper focuses on the challenging problem of 3D pose estimation of a diverse spectrum of articulated objects from single depth images.  ...  A novel structured prediction approach is considered, where 3D poses are represented as skeletal models that naturally operate on manifolds.  ...  The regression-based methods regard the prediction of 3D joint locations directly as a multivariate regression problem in 3D Euclidean space.  ... 
arXiv:1612.00596v1 fatcat:hld3xzsht5cfnlwfae2odliivi

Model-based Deep Hand Pose Estimation [article]

Xingyi Zhou, Qingfu Wan, Wei Zhang, Xiangyang Xue, Yichen Wei
2016 arXiv   pre-print
In this work, we propose a model based deep learning approach that adopts a forward kinematics based layer to ensure the geometric validity of estimated poses.  ...  For the first time, we show that embedding such a non-linear generative process in deep learning is feasible for hand pose estimation.  ...  Conclusions We show that it is possible to integrate the forward kinematic process of an articulated hand model into the deep learning framework for effective hand pose estimation.  ... 
arXiv:1606.06854v1 fatcat:tguioxut2fcs5mhlgmtp3gj2wq

Learning Body Shape and Pose from Dense Correspondences

Yusuke Yoshiyasu, Lucas Gamez
2020 Annual Conference of the European Association for Computer Graphics  
In this paper, we address the problem of learning 3D human pose and body shape from 2D image dataset, without having to use 3D supervisions (body shape and pose) which are in practice difficult to obtain  ...  To do so, we propose a training strategy called "deform-and-learn" where we alternate deformable surface registration and training of deep convolutional neural networks (ConvNets).  ...  obtained by deformable registration as annotations for training deep ConvNets, we regress body shape and pose parameters with an image.  ... 
doi:10.2312/egs.20201012 dblp:conf/eurographics/YoshiyasuG20 fatcat:ffbrncstkvauzj4clmyi76paem

Deep Kinematics Analysis for Monocular 3D Human Pose Estimation

Jingwei Xu, Zhenbo Yu, Bingbing Ni, Jiancheng Yang, Xiaokang Yang, Wenjun Zhang
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Above three steps are seamlessly integrated into deep neural models, which form a deep kinematics analysis pipeline concurrently considering the static/dynamic structure of 2D inputs and 3D outputs.  ...  For monocular 3D pose estimation conditioned on 2D detection, noisy/unreliable input is a key obstacle in this task.  ...  Conclusion In this paper, we propose a deep kinematics analysis framework for monocular 3D pose estimation.  ... 
doi:10.1109/cvpr42600.2020.00098 dblp:conf/cvpr/XuYNYY020 fatcat:kfqkv7wdnzblninha2pbe2lana

PMnet: Learning of Disentangled Pose and Movement for Unsupervised Motion Retargeting

Jongin Lim, Hyung Jin Chang, Jin Young Choi
2019 British Machine Vision Conference  
At each frame, to follow the pose of the input character, PMnet learns how to make the input pose first and then adjusts it to fit the target character's kinematic configuration.  ...  In this paper, we propose a deep learning framework for unsupervised motion retargeting.  ...  However, these methods can not be applied to motion retargeting because they require re-projection onto kinematic constraint to avoid invalid bone length configuration as they regress on joint positions  ... 
dblp:conf/bmvc/LimCC19 fatcat:e5qgxywyinhlna3cx7zj73srta

3D Human Pose Estimation on a Configurable Bed from a Pressure Image [article]

Henry M. Clever, Ariel Kapusta, Daehyung Park, Zackory Erickson, Yash Chitalia, Charles C. Kemp
2018 arXiv   pre-print
However, prior work on estimating human pose from pressure images has been restricted to 2D pose estimates and flat beds.  ...  The first network directly outputs 3D joint positions, while the second outputs a kinematic model that includes estimated joint angles and limb lengths.  ...  Fig. 7 shows the kinematics ConvNet with length regression adjusting for humans of different sizes and in different poses.  ... 
arXiv:1804.07873v2 fatcat:kzwntfkotbbldnfauel5vnb7zm

Deep Network for the Integrated 3D Sensing of Multiple People in Natural Images

Andrei Zanfir, Elisabeta Marinoiu, Mihai Zanfir, Alin-Ionut Popa, Cristian Sminchisescu
2018 Neural Information Processing Systems  
We design a multi-task deep neural network with differentiable stages where the person grouping problem is formulated as an integer program based on learned body part scores parameterized by both 2d and  ...  limbs) in images, groups them based on 2d and 3d information fused using learned scoring functions, and optimally aggregates such responses into partial or complete 3d human skeleton hypotheses under kinematic  ...  Given an image I, the processing stages are as follows: Deep Feature Extractor to compute features M I , Deep Volume Encoding to regress volumes containing 2d and 3d pose information V I .  ... 
dblp:conf/nips/ZanfirMZPS18 fatcat:5nw6slns6zh6xnqp6wivgdk234

Deep Unsupervised 3D Human Body Reconstruction from a Sparse set of Landmarks

Meysam Madadi, Hugo Bertiche, Sergio Escalera
2021 International Journal of Computer Vision  
Finally, a cascading network is applied to regress parameters of a statistical generative model that reconstructs body.  ...  In this paper we propose the first deep unsupervised approach in human body reconstruction to estimate body surface from a sparse set of landmarks, so called DeepMurf.  ...  Regress pose and shape parameters Our goal is to estimate SMPL pose and shape parameters from landmarks.  ... 
doi:10.1007/s11263-021-01488-2 fatcat:qtwhj5myevd3hj6yxfxqcbtyzm

SMPLR: Deep SMPL reverse for 3D human pose and shape recovery [article]

Meysam Madadi, Hugo Bertiche, Sergio Escalera
2019 arXiv   pre-print
In this paper we propose to embed SMPL within a deep model to accurately estimate 3D pose and shape from a still RGB image.  ...  We use CNN-based 3D joint predictions as an intermediate representation to regress SMPL pose and shape parameters. Later, 3D joints are reconstructed again in the SMPL output.  ...  The goal is to estimate SMPL parameters from image I using deep learning without directly regressing them.  ... 
arXiv:1812.10766v2 fatcat:ih33yr3dbndatk3lceqwh4heuy

Towards Kinematically Constrained Real Time Human Pose Estimation using Sparse IMUs

Deepak Nagaraj, Rhett Dobinson, Dirk Werth
2021 AAAI Spring Symposia  
This non-position paper deals with a hybrid approach involving full-body inverse kinematics (IK) and deep learning in order to estimate physiologically feasible joint angles in real time, based on orientation  ...  A bidirectional recurrent neural network (bi-RNN) is then trained using a newly collected IMU dataset to regress from the orientation data of 6 sensors to the joint angles obtained from IK.  ...  Figure 1 : 1 Figure 1: Proposed approach involving Inverse Kinematics and Deep Learning Figure 2 : 2 Figure 2: Convergence of MSE during training the bi-RNN Figure 3 : 3 Figure 3: Comparison of the  ... 
dblp:conf/aaaiss/NagarajDW21 fatcat:yyxgehenpnf5vidk5dtqbm4n5q

Model-based Hand Pose Estimation for Generalized Hand Shape with Appearance Normalization [article]

Jan Wöhlke, Shile Li, Dongheui Lee
2018 arXiv   pre-print
Recently, a hybrid approach has embedded a kinematic layer into the deep learning structure in such a way that the pose estimates obey the physical constraints of human hand kinematics.  ...  Since the emergence of large annotated datasets, state-of-the-art hand pose estimation methods have been mostly based on discriminative learning.  ...  [24] integrate the kinematic constraints into the loss of their human pose regression network. They use a bone-based instead of a joint-based representation of the pose.  ... 
arXiv:1807.00898v1 fatcat:qdoort3vpfbjtdcs6chq2esdvi

HMTNet:3D Hand Pose Estimation from Single Depth Image Based on Hand Morphological Topology [article]

Weiguo Zhou, Xin Jiang, Chen Chen, Sijia Mei, Yun-Hui Liu
2019 arXiv   pre-print
Next, regression module inspired from hand morphological topology is proposed.  ...  Thanks to the rapid development of CNNs and depth sensors, great progress has been made in 3D hand pose estimation.  ...  In the following, we will review the related works using hand model or hand kinematic constraints. A. HAND KINEMATIC CONSTRAINTS Deep-Prior [7] proposed by Oberweger is widely used recently.  ... 
arXiv:1911.04930v1 fatcat:2tn3v7z4bffhlkgwawchqdrefm

DeepHPS: End-to-end Estimation of 3D Hand Pose and Shape by Learning from Synthetic Depth [article]

Jameel Malik, Ahmed Elhayek, Fabrizio Nunnari, Kiran Varanasi, Kiarash Tamaddon, Alexis Heloir, Didier Stricker
2018 arXiv   pre-print
Then, a novel hand pose and shape layer, embedded inside our deep framework, produces 3D joint positions and hand mesh.  ...  In contrast to the existing methods which optimize only for joint positions, we propose a fully supervised deep network which learns to jointly estimate a full 3D hand mesh representation and pose from  ...  A novel deep network layer which performs: (a) Forward kinematics using a new combination of hand pose and bone scales parameters.  ... 
arXiv:1808.09208v1 fatcat:gwzkoqq6vne63gsxbrbbkisq74
« Previous Showing results 1 — 15 out of 5,731 results