Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
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 JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

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 journal

Trap tales: The influence of red alder stand conditions and forest fragmentation on family-level beetle bycatch diversity.

PloS one·2026
Same journal

MamNet-PT: A Mamba-enhanced hybrid architecture with selective state-space modeling for uncertainty-aware brain tumor segmentation.

PloS one·2026
Same journal

Multicenter evaluation of BACT-Info. and an infection algorithm using Urine Flow Cytometry among clinically diagnosed UTI patients in Indonesia.

PloS one·2026
Same journal

Cross-cultural adaptation and psychometric properties study of Prolonged Grief Disorder Questionnaire (PG-12-R) for caregivers of terminal cancer patients, Thai version.

PloS one·2026
Same journal

Design and in silico validation of donor DNA for RNA-guided recombinase-mediated knockout of mstnb gene in Labeo rohita.

PloS one·2026
Same journal

ViT-MultiRAGNet: A scalable and reliable retrieval-augmented Vision Transformer framework for memory-guided feature fusion multi-modal mammogram classification.

PloS one·2026

相关实验视频

Updated: Sep 18, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

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

Published on: December 15, 2023

650

PoseNet++:一个多尺度和优化的特征提取网络,用于高精度的人体姿势估计.

Chao Lv1, Geyao Ma1

  • 1College of Electronic Information Engineering, Changchun University of Science and Technology, Changchun, China.

PloS one
|June 25, 2025
PubMed
概括

PoseNet++显著提高了人体姿势估计的准确性,特别是在具有挑战性的场景中,如遮蔽和多人设置. 这种新的深度学习方法可以在模型复杂度降低的情况下获得最先进的结果.

科学领域:

  • 计算机视觉 计算机视觉
  • 深度学习 (Deep Learning) 是一种深度学习.
  • 人工智能的人工智能

背景情况:

  • 深度学习已经推进了人类姿势估计 (HPE),但仍然存在阻塞,复杂姿势和多人互动的挑战.
  • 现有的模型在混乱或部分模糊的场景中难以准确识别身体关节的关键位置.

研究的目的:

  • 开发一种新的人类姿势估计方法,PoseNet++,可以克服精度和效率的局限性.
  • 通过创新的建筑模块,增强模型处理遮,复杂姿势和多人场景的能力.

主要方法:

  • 推出了PoseNet++,一个三叠的沙钟架构,具有三个关键的创新:多尺度空间金字塔注意力沙钟模块 (MSPAHM),坐标通道先前卷积注意力 (C-CPCA) 和PinSK瓶剩余模块 (PBRM).
  • MSPAHM 改进了远程通道依赖性,以更好地捕捉闭塞下的联合关系.
  • C-CPCA优先考虑关键点区域并减少多人设置中的混乱,而PBRM优化了复杂姿势的特征提取.

主要成果:

  • 与基线相比,PoseNet++在MPII验证集上的PCKh得分取得了3.3%的相对改善.
  • 该模型显著减少了参数60.3%,浮点运算减少了53.1%.
  • 在较低模型复杂度的MPII,LSP,COCO和CrowdPose数据集上实现了最先进的性能.

更多相关视频

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
09:41

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

Published on: April 21, 2023

1.7K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

530

相关实验视频

Last Updated: Sep 18, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

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

Published on: December 15, 2023

650
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
09:41

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

Published on: April 21, 2023

1.7K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

530

结论:

  • PoseNet++为人类姿势估计提供了一个高度准确和高效的解决方案.
  • 提出的架构创新有效地解决了HPE的关键挑战,包括阻塞和复杂的场景.
  • PoseNet++代表了基于深度学习的人类姿势估计的重大进步,平衡了性能和计算成本.