Complex Human Pose Estimation via Keypoints Association Constraint Network

Zhu, Xuan; Guo, Zhenpeng; Liu, Xin; Li, Bin; Peng, Jinye; Chen, Peirong; Wang, Rongzhi

doi:10.1109/access.2020.3037736

Cited by 6 publications

(4 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [10], Zhu et al, present the key-point association constraint network (KACNet), which is a novel network for estimating human posture. For this, Channel-1 of this network is used to determine the location of key points where the distance loss function is a constraint, and channel-2 of the network is restricted to obtain the link between key points using the association loss function.…”

Section: B Multi Person Pose Estimationmentioning

confidence: 99%

See 1 more Smart Citation

Multi-Person Pose Estimation Using Group-Based Convolutional Neural Network Model

et al. 2023

View full text Add to dashboard Cite

Human pose estimation has drawn extensive attention recently and there has been significant progress on it due to the rising popularity of convolutional neural networks (CNN). However, existing stateof-the-art approaches suffer from occlusion, complicated backgrounds, and substantial position fluctuations because of disregarding the human body form. Human parsing is a very pertinent activity that can provide crucial semantic data about bodily parts for position estimation. To overcome the aforesaid limitations, this paper introduces a human pose estimation method using a group-based convolutional neural network model. The proposed method adopts a bottom-up parsing strategy that yields features to extract skeletal key points in the human body. Moreover, it creates a grouping of anatomical key points for individuals by utilizing the non-parametric description for the key point association vector field. Experimental results indicate that the proposed method provides superior performance than the state-of-the-art algorithms in terms of accuracy. In addition, it optimizes its output and detects occluded as well as invisible key points by incorporating feature representation. The proposed method surpasses the recent methods, achieving 93% of the mean average accuracy.

show abstract

Section: B Multi Person Pose Estimationmentioning

confidence: 99%

“…Besides, the runtime is directly correlated with the population of the image [7], [8], [9]. Conversely, bottom-up strategies are desirable because they give stability to early investment and have the capacity to dissociate runtime challenge based on the number of people in the picture [3], [4], [10], [11].…”

Section: Introductionmentioning

confidence: 99%

Multi-Person Pose Estimation Using Group-Based Convolutional Neural Network Model

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Step 2: obtain the ðx z , y z Þ of multiple similar human bodies through the nonmaximum suppression method. After locating a probability peak point, this method can control the probability values of other fields to be lower than this value, so as to avoid repeated detection of the posture ðx z , y z Þ of the same human body [21].…”

Section: Sports Posture Estimation Based On Cnnmentioning

confidence: 99%

Accurate Recognition Method of Continuous Sports Action Based on Deep Learning Algorithm

Zhengfeng

2022

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

In order to improve the recognition effect of continuous sports action, an accurate recognition method of continuous sports action based on deep learning is proposed. The convolution neural network is used to predict the key points and center vector of human body in grid mode, and the key points of human body are grouped according to the distance from the key points of the human body to the center of human body, so as to obtain the sequence diagram of sports posture, input the posture sequence diagram of sports action in the deep learning network, and apply the best strategy of continuous action recognition obtained in the deep learning training stage to the accurate recognition of continuous sports action. The position of the rectangular frame of the attitude instead of the target attitude is predicted online by the convolution neural network, and the recognition is completed when the attitude stops. The experimental results show that this method can effectively recognize continuous sports actions in real time. The expected average overlap rate of different video attributes is high, which has better recognition effect. At different position error thresholds, it has high recognition accuracy.

show abstract

“…Human pose estimation is a way of identifying and classifying the human body’s joints in images. It generally uses the keypoint estimation method to select a set of most representative points in human pose [ 33 , 34 ], such as head, shoulders, elbows, wrists, hips, knees, ankles, and portray the human pose by connecting the lines. However, identifying these joints, especially manually, requires rich texture information.…”

Section: Introductionmentioning

confidence: 99%

Identity-Preserved Human Posture Detection in Infrared Thermal Images: A Benchmark

Guo

Chen

Deng

et al. 2022

Sensors

View full text Add to dashboard Cite

Human pose estimation has a variety of real-life applications, including human action recognition, AI-powered personal trainers, robotics, motion capture and augmented reality, gaming, and video surveillance. However, most current human pose estimation systems are based on RGB images, which do not seriously take into account personal privacy. Although identity-preserved algorithms are very desirable when human pose estimation is applied to scenarios where personal privacy does matter, developing human pose estimation algorithms based on identity-preserved modalities, such as thermal images concerned here, is very challenging due to the limited amount of training data currently available and the fact that infrared thermal images, unlike RGB images, lack rich texture cues which makes annotating training data itself impractical. In this paper, we formulate a new task with privacy protection that lies between human detection and human pose estimation by introducing a benchmark for IPHPDT (i.e., Identity-Preserved Human Posture Detection in Thermal images). This task has a threefold novel purpose: the first is to establish an identity-preserved task with thermal images; the second is to achieve more information other than the location of persons as provided by human detection for more advanced computer vision applications; the third is to avoid difficulties in collecting well-annotated data for human pose estimation in thermal images. The presented IPHPDT dataset contains four types of human postures, consisting of 75,000 images well-annotated with axis-aligned bounding boxes and postures of the persons. Based on this well-annotated IPHPDT dataset and three state-of-the-art algorithms, i.e., YOLOF (short for You Only Look One-level Feature), YOLOX (short for Exceeding YOLO Series in 2021) and TOOD (short for Task-aligned One-stage Object Detection), we establish three baseline detectors, called IPH-YOLOF, IPH-YOLOX, and IPH-TOOD. In the experiments, three baseline detectors are used to recognize four infrared human postures, and the mean average precision can reach 70.4%. The results show that the three baseline detectors can effectively perform accurate posture detection on the IPHPDT dataset. By releasing IPHPDT, we expect to encourage more future studies into human posture detection in infrared thermal images and draw more attention to this challenging task.

show abstract

Complex Human Pose Estimation via Keypoints Association Constraint Network

Cited by 6 publications

References 40 publications

Multi-Person Pose Estimation Using Group-Based Convolutional Neural Network Model

Multi-Person Pose Estimation Using Group-Based Convolutional Neural Network Model

Accurate Recognition Method of Continuous Sports Action Based on Deep Learning Algorithm

Identity-Preserved Human Posture Detection in Infrared Thermal Images: A Benchmark

Contact Info

Product

Resources

About