Self-Supervised Learning for Human Pose Estimation in Sports

Ludwig, Katja; Scherer, Sebastian; Einfalt, Moritz; Lienhart, Rainer

doi:10.1109/icmew53276.2021.9456000

Cited by 17 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As most datasets annotate key points after the K th frames, leaving some frames without annotation, several studies [8,112,113] have relied on unsupervised approaches to estimate key points along video frames. Zhang et al [112] introduced the Key Frame Proposal Network (K-FPN), which recovers the entire pose sequence unsupervised by selecting spatial and temporal information from a set of keyframes.…”

Section: Single Pose Estimation Video-basedmentioning

confidence: 99%

See 1 more Smart Citation

Human Pose Estimation Using Deep Learning: A Systematic Literature Review

Samkari,

Arif,

Alghamdi

et al. 2023

MAKE

View full text Add to dashboard Cite

Human Pose Estimation (HPE) is the task that aims to predict the location of human joints from images and videos. This task is used in many applications, such as sports analysis and surveillance systems. Recently, several studies have embraced deep learning to enhance the performance of HPE tasks. However, building an efficient HPE model is difficult; many challenges, like crowded scenes and occlusion, must be handled. This paper followed a systematic procedure to review different HPE models comprehensively. About 100 articles published since 2014 on HPE using deep learning were selected using several selection criteria. Both image and video data types of methods were investigated. Furthermore, both single and multiple HPE methods were reviewed. In addition, the available datasets, different loss functions used in HPE, and pretrained feature extraction models were all covered. Our analysis revealed that Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) are the most used in HPE. Moreover, occlusion and crowd scenes remain the main problems affecting models’ performance. Therefore, the paper presented various solutions to address these issues. Finally, this paper highlighted the potential opportunities for future work in this task.

show abstract

Section: Single Pose Estimation Video-basedmentioning

confidence: 99%

“…For example, in Hajj and Umrah, human poses are estimated to detect abnormal behavior when people throw pebbles or perform Tawaf [3]. Other applications of HPE include interacting with virtual worlds [4], animating game characters [5,6], tracking patient movements [7], and analyzing sports performance [8].…”

Section: Introductionmentioning

confidence: 99%

Human Pose Estimation Using Deep Learning: A Systematic Literature Review

Samkari,

Arif,

Alghamdi

et al. 2023

MAKE

View full text Add to dashboard Cite

show abstract

“…A lot of methods address human motion analysis in sports videos, differently. Based on self-supervised learning, the suggested method in [52] employs pseudo labels as a selfsupervised training strategy, together with a pseudo label filtering method, in the disciplines of triple and long jump.…”

Section: E Sports Posementioning

confidence: 99%

DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training

et al. 2023

View full text Add to dashboard Cite

The tremendous progress of deep convolution neural networks has shown promising results on the classification of various sports activities. However, the accurate localization of a particular sports event or activity in a continuous video stream is still a challenging problem. The accurate detection of sports actions enables the comparison of different performances, objectively. In this work, we propose the DiveNet action localization module to detect the springboard diving sports action in an unconstrained environment. We used Temporal Convolution Network (TCN) over a backbone feature extractor to localize diving actions, with low latency. We estimate the divers center of mass (COM) trajectory and the peak dive height using the temporal demarcations provided by the action localization step via the projectile motion formula. In addition, we train a DiveNet pose regression network, which extends the Unipose architecture with direct physical parameter estimation, i.e COM and 2D joint keypoints. We propose a new homography computation method between the diving motion plane and the image-view for each dive. This enables the representation of physical parameters in metric scale, without any calibration. We release the first publicly available diving sports video dataset, recorded at 60 Hz with a static camera setup for different springboard heights. DiveNet action localization achieves an accuracy of 95% with a single frame latency (< 25 ms). The DiveNet pose regression model shows competitive results around 70% PCK on different diving pose datasets. We achieve COM accuracy of 6 pixels, dive peak height sensitivity of 20 cm and mean joint angle errors around 10 degrees. INDEX TERMSdeep learning; diving sports pose; action localization; sports analytics; I. INTRODUCTION

show abstract

“…Other emerging techniques, such as semi‐supervised, self‐supervised learning, contrastive learning (Koshkina et al, 2021), and so on, are gaining immense popularity among researchers. In recent studies, Ludwig et al (2021) adopted self‐supervised learning to learn the feature representation from the unlabeled images and used it to estimate the 2D human pose for long and triple jump sports. They utilized two methods of self‐supervised a mean teacher approach and generating pseudo labels from the unlabeled images.…”

Section: Challenges and Emerging Technologiesmentioning

confidence: 99%

Sports analytics review: Artificial intelligence applications, emerging technologies, and algorithmic perspective

Ghosh

Ramamurthy

Chakma

et al. 2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

The rapid and impromptu interest in the coupling of machine learning (ML) algorithms with wearable and contactless sensors aimed at tackling real‐world problems warrants a pedagogical study to understand all the aspects of this research direction. Considering this aspect, this survey aims to review the state‐of‐the‐art literature on ML algorithms, methodologies, and hypotheses adopted to solve the research problems and challenges in the domain of sports. First, we categorize this study into three main research fields: sensors, computer vision, and wireless and mobile‐based applications. Then, for each of these fields, we thoroughly analyze the systems that are deployable for real‐time sports analytics. Next, we meticulously discuss the learning algorithms (e.g., statistical learning, deep learning, reinforcement learning) that power those deployable systems while also comparing and contrasting the benefits of those learning methodologies. Finally, we highlight the possible future open‐research opportunities and emerging technologies that could contribute to the domain of sports analytics.This article is categorized under: Technologies > Machine Learning Technologies > Artificial Intelligence Technologies > Internet of Things

show abstract

Self-Supervised Learning for Human Pose Estimation in Sports

Cited by 17 publications

References 15 publications

Human Pose Estimation Using Deep Learning: A Systematic Literature Review

Human Pose Estimation Using Deep Learning: A Systematic Literature Review

DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training

Sports analytics review: Artificial intelligence applications, emerging technologies, and algorithmic perspective

Contact Info

Product

Resources

About