Search research articles

关于 JoVE

概览领导团队博客 JoVE 帮助中心

作者

出版流程编辑委员会范围与政策同行评审常见问题投稿

图书馆员

用户评价订阅访问资源图书馆顾问委员会常见问题

研究

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments 存档

教育

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual 教师资源中心教师网站

使用条款与条件

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序

Same author

Research Advances in the Pathogenesis of Sepsis-Associated Encephalopathy.

International journal of molecular sciences·2026

Same author

CoShMDM: Contact and Shape-Aware Latent Motion Diffusion Model for Human Interaction Generation.

IEEE transactions on visualization and computer graphics·2026

Same author

Cross-view contrastive representation learning on meta-path induced graphs with node features for bundle recommendation.

Neural networks : the official journal of the International Neural Network Society·2026

Same author

Development and validation of an interpretable machine learning model for early prediction in patients with diabetes and sepsis.

Scientific reports·2025

Same author

Universal and highly sensitive detection of influenza A virus and streptococcus pneumoniae using WGA-modified magnetic SERS nanotags-based lateral flow assay.

Nanomedicine : nanotechnology, biology, and medicine·2025

Same author

Domain-aware self-prompting for cross-domain sequential recommendations with natural language explanations.

Neural networks : the official journal of the International Neural Network Society·2025

Search research articles

相关实验视频

Updated: Apr 12, 2026

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Published on: April 21, 2023

FingerPoseNet:一个指数级多任务学习网络,用于3D手姿势估计的残余特征共享.

Tekie Tsegay Tewolde¹, Ali Asghar Manjotho¹, Prodip Kumar Sarker²

¹School of Computer Science and Technology, Beijing Institute of Technology, Beijing, 100081, China.

Neural networks : the official journal of the International Neural Network Society

|March 13, 2025

概括

FingerPoseNet通过专注于手指水平特征来增强3D手姿势估计. 这种新的方法提高了从深度图像中捕捉手臂关节的准确性.

关键词:

手的姿势估计手的姿势估计信息共享信息共享多任务学习多任务学习用户行为建模.虚拟现实虚拟现实就是虚拟现实.

更多相关视频

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Utilizing vmTracking to Improve the Accuracy of Multi-Animal Pose Estimation in Rodent Social Behavior Studies

Utilizing vmTracking to Improve the Accuracy of Multi-Animal Pose Estimation in Rodent Social Behavior Studies

Published on: November 7, 2025

相关实验视频

Last Updated: Apr 12, 2026

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Published on: April 21, 2023

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Utilizing vmTracking to Improve the Accuracy of Multi-Animal Pose Estimation in Rodent Social Behavior Studies

Utilizing vmTracking to Improve the Accuracy of Multi-Animal Pose Estimation in Rodent Social Behavior Studies

Published on: November 7, 2025

科学领域:

计算机视觉计算机视觉
机器学习机器学习
机器人技术机器人技术机器人技术

背景情况:

当前的3D手姿势估计方法经常使用共享的特征图,限制了关键的手指水平细节的增强.
准确的关节到手指的关联和关节对于精确的手姿势估计至关重要,但使用现有技术捕获这些关联和关节是具有挑战性的.

研究的目的:

介绍FingerPoseNet,一个新的指数级多任务学习网络,用于从深度图像中准确的3D手姿势估计.
解决当前方法在增强手指水平特征的局限性,以改进手的关节分析.

主要方法:

FingerPoseNet采用三级架构:一个ResNet-50骨干用于共享特征提取,一个指级多任务学习阶段用于增强个别的手指和手掌特征,以及一个多任务融合层.
采用多任务学习,将手姿势估计分解为六个子任务 (每个手指和手掌一个),每个处理特征提取,增强和3D关键点回归.
引入剩余功能共享方法,在所有子任务中挖掘补充信息,增强子任务特定的功能.

主要成果:

与最先进的方法相比,FingerPoseNet在准确度方面取得了显著的改进.
在五个具有挑战性的公共数据集 (ICVL,NYU,MSRA,Hands-2019-Task1,HO3D-v3) 上进行的实验验证实了拟议方法的有效性.

结论:

FingerPoseNet有效地解决了在3D手姿势估计中增强手指水平特征的挑战.
拟议的指级多任务学习网络与残余特征共享提供了一个强大而准确的解决方案,用于从深度数据中估计3D手姿势.