Filters








1,129 Hits in 4.7 sec

3D Human Pose Estimation with 2D Marginal Heatmaps [article]

Aiden Nibali, Zhen He, Stuart Morgan, Luke Prendergast
2018 arXiv   pre-print
We are also able to achieve state-of-the-art accuracy on publicly available 3D human pose estimation data.  ...  We propose improvements to 3D coordinate prediction which avoid the aforementioned undesirable traits by predicting 2D marginal heatmaps under an augmented soft-argmax scheme.  ...  Conclusion 2D marginal heatmaps are a memory-efficient alternative to volumetric heatmaps when building models for 3D pose estimation.  ... 
arXiv:1806.01484v2 fatcat:d5nlhqqk6vglze2lfavq5rbvwq

I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image [article]

Gyeongsik Moon, Kyoung Mu Lee
2020 arXiv   pre-print
Most of the previous image-based 3D human pose and mesh estimation methods estimate parameters of the human mesh model from an input image.  ...  Our lixel-based 1D heatmap preserves the spatial relationship in the input image and models the prediction uncertainty.  ...  Heatmap-based 3D human pose estimation.  ... 
arXiv:2008.03713v2 fatcat:h3lm7qdcj5cqze4dm2pwrwnhvu

Harvesting Multiple Views for Marker-less 3D Human Pose Annotations [article]

Georgios Pavlakos, Xiaowei Zhou, Konstantinos G. Derpanis, Kostas Daniilidis
2017 arXiv   pre-print
Starting from a generic ConvNet for 2D human pose, and assuming a multi-view setup, we describe an automatic way to collect accurate 3D human pose annotations.  ...  ,"personalization"), and (ii) training a ConvNet from scratch for single view 3D human pose prediction without leveraging 3D pose groundtruth.  ...  Given a multi-view camera setup, we use a generic ConvNet for 2D human pose estimation [27] , and produce single-view pose predictions in the form of 2D heatmaps for each view.  ... 
arXiv:1704.04793v1 fatcat:vs6s6bb7mvgzfci7gysq5p477a

Harvesting Multiple Views for Marker-Less 3D Human Pose Annotations

Georgios Pavlakos, Xiaowei Zhou, Konstantinos G. Derpanis, Kostas Daniilidis
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Starting from a generic ConvNet for 2D human pose, and assuming a multi-view setup, we describe an automatic way to collect accurate 3D human pose annotations.  ...  ,"personalization"), and (ii) training a ConvNet from scratch for single view 3D human pose prediction without leveraging 3D pose groundtruth.  ...  Given a multi-view camera setup, we use a generic ConvNet for 2D human pose estimation [27] , and produce single-view pose predictions in the form of 2D heatmaps for each view.  ... 
doi:10.1109/cvpr.2017.138 dblp:conf/cvpr/PavlakosZDD17 fatcat:gatlbmqhdzah5hfzejcqwpqxxq

Probabilistic Monocular 3D Human Pose Estimation with Normalizing Flows [article]

Tom Wehrbein, Marco Rudolph, Bodo Rosenhahn, Bastian Wandt
2021 arXiv   pre-print
3D human pose estimation from monocular images is a highly ill-posed problem due to depth ambiguities and occlusions.  ...  To this end, we propose a normalizing flow based method that exploits the deterministic 3D-to-2D mapping to solve the ambiguous inverse 2D-to-3D problem.  ...  In contrast, we model monocular 3D human pose estimation as an ambiguous inverse problem with multiple feasible solutions.  ... 
arXiv:2107.13788v2 fatcat:twdcqxdf3zbfhlpnpi4cybldqi

Revisiting Skeleton-based Action Recognition [article]

Haodong Duan, Yue Zhao, Kai Chen, Dahua Lin, Bo Dai
2022 arXiv   pre-print
In this work, we propose PoseC3D, a new approach to skeleton-based action recognition, which relies on a 3D heatmap stack instead of a graph sequence as the base representation of human skeletons.  ...  Compared to GCN-based methods, PoseC3D is more effective in learning spatiotemporal features, more robust against pose estimation noises, and generalizes better in cross-dataset settings.  ...  (a) 2D poses estimated with HRNet. (b) 3D poses collected with Kinect. (c) 3D poses estimated with VIBE. Figure 1 . 1 Figure 1. PoseConv3D takes 2D poses as inputs.  ... 
arXiv:2104.13586v2 fatcat:b64skl5xtffy3c22awbdhgxh3m

Cross View Fusion for 3D Human Pose Estimation [article]

Haibo Qiu, Chunyu Wang, Jingdong Wang, Naiyan Wang, Wenjun Zeng
2019 arXiv   pre-print
It consists of two separate steps: (1) estimating the 2D poses in multi-view images and (2) recovering the 3D poses from the multi-view 2D poses.  ...  Second, we present a recursive Pictorial Structure Model to recover the 3D pose from the multi-view 2D poses. It gradually improves the accuracy of 3D pose with affordable computational cost.  ...  [18] extend PSM to multi-view 3D human pose estimation. For example, in [18] , they first estimate 2D poses independently for each view and then recover the 3D pose using PSM.  ... 
arXiv:1909.01203v1 fatcat:wcfiuoyakrd3bpjsle2dq37pve

Synthetic Occlusion Augmentation with Volumetric Heatmaps for the 2018 ECCV PoseTrack Challenge on 3D Human Pose Estimation [article]

István Sárándi, Timm Linder, Kai O. Arras, Bastian Leibe
2018 arXiv   pre-print
In this paper we present our winning entry at the 2018 ECCV PoseTrack Challenge on 3D human pose estimation.  ...  Absolute person center depth is estimated by a 1D heatmap prediction head. The coordinates are back-projected to 3D camera space, where we minimize the L1 loss.  ...  Recently, building on experience gained from 2D human pose estimation (e.g., [16] ), heatmap-like methods have been introduced for 3D pose es- timation with promising results.  ... 
arXiv:1809.04987v3 fatcat:icrbfpuok5egtmxjzumynvv3aa

HeatER: An Efficient and Unified Network for Human Reconstruction via Heatmap-based TransformER [article]

Ce Zheng, Matias Mendieta, Taojiannan Yang, Chen Chen
2022 arXiv   pre-print
Recently, vision transformers have shown great success in 2D human pose estimation (2D HPE), 3D human pose estimation (3D HPE), and human mesh reconstruction (HMR) tasks.  ...  A heatmap reconstruction module is applied to improve the robustness of the estimated human pose and mesh.  ...  Heatmap Reconstruction Module Compared with estimating 2D human pose, recovering 3D pose and mesh are more challenging due to depth ambiguity and occlusion.  ... 
arXiv:2205.15448v1 fatcat:2slxbzubg5a3hlakrjfln3xttq

HEMlets Pose: Learning Part-Centric Heatmap Triplets for Accurate 3D Human Pose Estimation [article]

Kun Zhou, Xiaoguang Han, Nianjuan Jiang, Kui Jia, Jiangbo Lu
2019 arXiv   pre-print
Estimating 3D human pose from a single image is a challenging task.  ...  the 2D observation and the 3D interpretation.  ...  [18] proposed fusing 2D joint locations with high-level image features to boost the estimation of 3D human pose.  ... 
arXiv:1910.12032v1 fatcat:zkjqmjmfmnh4timtxd6k7f7n6u

Fusing Wearable IMUs with Multi-View Images for Human Pose Estimation: A Geometric Approach [article]

Zhe Zhang, Chunyu Wang, Wenhu Qin, Wenjun Zeng
2020 arXiv   pre-print
We propose to estimate 3D human pose from multi-view images and a few IMUs attached at person's limbs.  ...  Then we lift the multi-view 2D poses to the 3D space by an Orientation Regularized Pictorial Structure Model (ORPSM) which jointly minimizes the projection error between the 3D and 2D poses, along with  ...  In the second step, we estimate 3D pose from multi-view 2D poses (heatmaps) by a Pictorial Structure Model (PSM) [12, 17, 2] .  ... 
arXiv:2003.11163v2 fatcat:24ml35kljbhudened2wr6xhkzu

Forecasting Human Dynamics from Static Images

Yu-Wei Chao, Jimei Yang, Brian Price, Scott Cohen, Jia Deng
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Our 3D-PFNet integrates recent advances on single-image human pose estimation and sequence prediction, and converts the 2D predictions into 3D space.  ...  We demonstrate competitive performance of our 3D-PFNet on 2D pose forecasting and 3D pose recovery through quantitative and qualitative results.  ...  Our proposed 3D Pose Forecasting Network (3D-PFNet) integrates recent advances on single-image human pose estimation and sequence prediction, and further converts the 2D predictions into 3D space.  ... 
doi:10.1109/cvpr.2017.388 dblp:conf/cvpr/ChaoYPCD17 fatcat:padjlbnk35hr5noubklmireszy

Hand Pose Estimation via Latent 2.5D Heatmap Regression [chapter]

Umar Iqbal, Pavlo Molchanov, Thomas Breuel, Juergen Gall, Jan Kautz
2018 Lecture Notes in Computer Science  
Estimating the 3D pose of a hand is an essential part of human-computer interaction.  ...  Our system achieves state-of-the-art accuracy for 2D and 3D hand pose estimation on several challenging datasets in presence of severe occlusions.  ...  The problem, however, shares several properties with human body pose estimation and many approaches proposed for the human body can be easily adapted for hand pose estimation.  ... 
doi:10.1007/978-3-030-01252-6_8 fatcat:v3qp4leaire3ra2bywz7z5ab7y

Recognizing Human Actions as the Evolution of Pose Estimation Maps

Mengyuan Liu, Junsong Yuan
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
With recent advances of human pose estimation, this work presents a novel method to recognize human action as the evolution of pose estimation maps.  ...  Specifically, the evolution of pose estimation maps can be decomposed as an evolution of heatmaps, e.g., probabilistic maps, and an evolution of estimated 2D human poses, which denote the changes of body  ...  The success of 3D human pose inspires us to estimate 2D human poses from videos for action recognition.  ... 
doi:10.1109/cvpr.2018.00127 dblp:conf/cvpr/LiuY18 fatcat:i3rmf6vm4jh3zhhouxvzwi5a5i

Forecasting Human Dynamics from Static Images [article]

Yu-Wei Chao, Jimei Yang, Brian Price, Scott Cohen, Jia Deng
2017 arXiv   pre-print
Our 3D-PFNet integrates recent advances on single-image human pose estimation and sequence prediction, and converts the 2D predictions into 3D space.  ...  We demonstrate competitive performance of our 3D-PFNet on 2D pose forecasting and 3D pose recovery through quantitative and qualitative results.  ...  between 2D heatmaps and 3D ground truth of {∆, T, f }.  ... 
arXiv:1704.03432v1 fatcat:4l4jlzuc6jgvdajcfwqxpdhccq
« Previous Showing results 1 — 15 out of 1,129 results