Single-Stage Multi-Person Pose Machines

Nie, Xuecheng; Feng, Jiashi; Zhang, Jianfeng; Yan, Shuicheng

doi:10.1109/iccv.2019.00705

Cited by 230 publications

(140 citation statements)

References 33 publications

Supporting

Mentioning

139

Contrasting

Order By: Relevance

“…Recently, multi-pers on pose estimation has aroused a great interest due to the real-life demand. Nowadays, researchers have made painstaking efforts [17,29,38,40] to accelerate its progress. For examples, CASNet [38] improves the feature representation via adopting the spatial and channel-wise attention.…”

Section: Related Workmentioning

confidence: 99%

Alleviating Human-level Shift

Xu¹,

Zou²,

Lin³

2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Human pose estimation has been widely studied with much focus on supervised learning requiring sufficient annotations. However, in real applications, a pretrained pose estimation model usually need be adapted to a novel domain with no labels or sparse labels. Such domain adaptation for 2D pose estimation hasn't been explored. The main reason is that a pose, by nature, has typical topological structure and needs fine-grained features in local keypoints. While existing adaptation methods do not consider topological structure of object-of-interest and they align the whole images coarsely. Therefore, we propose a novel domain adaptation method for multi-person pose estimation to conduct the human-level topological structure alignment and fine-grained feature alignment. Our method consists of three modules: Cross-Attentive Feature Alignment (CAFA), Intra-domain Structure Adaptation (ISA) and Inter-domain Human-Topology Alignment (IHTA) module. The CAFA adopts a bidirectional spatial attention module (BSAM) that focuses on fine-grained local feature correlation between two humans to adaptively aggregate consistent features for adaptation. We adopt ISA only in semi-supervised domain adaptation (SSDA) to exploit the corresponding keypoint semantic relationship for reducing the intra-domain bias. Most importantly, we propose an IHTA to learn more domain-invariant human topological representation for reducing the inter-domain discrepancy. We model the human topological structure via the graph convolution network (GCN), by passing messages on which, high-order relations can be considered. This structure preserving alignment based on GCN is beneficial to the occluded or extreme pose inference. Extensive experiments are conducted on two popular benchmarks and results demonstrate the competency of our method compared with existing supervised approaches. CCS CONCEPTS • Computing methodologies → Activity recognition and understanding; Transfer learning;

show abstract

Section: Related Workmentioning

confidence: 99%

Alleviating Human-level Shift

Xu¹,

Zou²,

Lin³

2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…Recently, Nie et al [19] proposed a one-stage multi-person pose estimation method (SPM) which predicts root joints (person centroids) and joint displacements directly. Although both SPM and our method predict person centroid, they use centroid plus displacements to recover joints and we use centroid to guide joint grouping.…”

Section: Multi-person Pose Estimationmentioning

confidence: 99%

“…surveillance, autonomous driving, human-computer interaction, etc. In the last few years, considerable progress on human pose estimation has been achieved by deep learning based approaches [17,34,29,18,19,35].…”

Section: Introductionmentioning

confidence: 99%

Efficient Human Pose Estimation with Depthwise Separable Convolution and Person Centroid Guided Joint Grouping

2020

Pattern Recognition and Computer Vision

View full text Add to dashboard Cite

In this paper, we propose efficient and effective methods for 2D human pose estimation. A new ResBlock is proposed based on depthwise separable convolution and is utilized instead of the original one in Hourglass network. It can be further enhanced by replacing the vanilla depthwise convolution with a mixed depthwise convolution. Based on it, we propose a bottom-up multi-person pose estimation method. A rooted tree is used to represent human pose by introducing person centroid as the root which connects to all body joints directly or hierarchically. Two branches of sub-networks are used to predict the centroids, body joints and their offsets to their parent nodes. Joints are grouped by tracing along their offsets to the closest centroids. Experimental results on the MPII human dataset and the LSP dataset show that both our singleperson and multi-person pose estimation methods can achieve competitive accuracies with low computational costs.

show abstract

“…In contrast, the bottom-up strategy [7,8,9,10,11] predicts all body joints in a whole image at first, and then tries to group them to corresponding person instances. Comparing with the top-down methods, they avoid higher joint detection and better robustness to the increased number of persons in the image.…”

Section: Introductionmentioning

confidence: 99%

Partitioned Centerpose Network for Bottom-Up Multi-Person Pose Estimation

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

In bottom-up multi-person pose estimation method, grouping joint candidates into corresponding person instance is a challenging problem. In this paper, a new bottom-up method, Partitioned CenterPose (PCP) Network, is proposed to better cluster all detected joints. To achieve this goal, a novel Partition Pose Representation (PPR) is proposed which integrate person instance and body joint by joint offset. PPR leverages the center of human body and the offset between center point and body joint to encode human pose. To better enhance the relationship of body joints, we divide human body into five parts, and generate sub-PPR in each part. Based on PPR, PCP Network can detect persons and body joints simultaneously, and then grouping all body joints by joint offset. Moreover, an improved ℓ loss is designed to obtain more accurate joint offset. On the COCO keypoints dataset, the proposed method performs on par with the existing state-of-the-art bottom-up method in accuracy and speed.

show abstract

Single-Stage Multi-Person Pose Machines

Cited by 230 publications

References 33 publications

Alleviating Human-level Shift

Alleviating Human-level Shift

Efficient Human Pose Estimation with Depthwise Separable Convolution and Person Centroid Guided Joint Grouping

Partitioned Centerpose Network for Bottom-Up Multi-Person Pose Estimation

Contact Info

Product

Resources

About