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

Parallel Processing01:20

Parallel Processing

The brain processes sensory information rapidly due to parallel processing, which involves sending data across multiple neural pathways at the same time. This method allows the brain to manage various sensory qualities, such as shapes, colors, movements, and locations, all concurrently. For instance, when observing a forest landscape, the brain simultaneously processes the movement of leaves, the shapes of trees, the depth between them, and the various shades of green. This enables a quick and...

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

GoP-Based Quality Enhancement on Video Compression.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same author

Correction: Wu et al. Determination of Luteolin-7-O-diglucuronide in <i>Perilla frutescens</i> (L.) Britt. Leaf Extracts from Different Regions of China and Republic of Korea and Its Cholesterol-Lowering Effect. <i>Molecules</i> 2023, <i>28</i>, 7007.

Molecules (Basel, Switzerland)·2025
Same author

Masked Feature Residual Coding for Neural Video Compression.

Sensors (Basel, Switzerland)·2025
Same author

Sparse-DeRF: Deblurred Neural Radiance Fields From Sparse View.

IEEE transactions on pattern analysis and machine intelligence·2025
Same author

Recurrent Flow Update Model Using Image Pyramid Structure for 4K Video Frame Interpolation.

Sensors (Basel, Switzerland)·2025
Same author

Multi-Granularity Aggregation with Spatiotemporal Consistency for Video-Based Person Re-Identification.

Sensors (Basel, Switzerland)·2024

相关实验视频

Updated: Jun 23, 2026

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

9.0K

使用伪标签的多任务学习策略:面部识别,面部地标检测和头部姿势估计.

Yongju Lee1, Sungjun Jang1, Han Byeol Bae2

  • 1School of Electrical and Electronic Engineering, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 03722, Republic of Korea.

Sensors (Basel, Switzerland)
|May 25, 2024
PubMed
概括

这项研究引入了一种新的伪标签技术和多任务学习框架,以改善现实世界条件下的面部分析. 该方法增强了面部地标检测,头部姿势估计和面部识别,实现了最先进的性能.

关键词:
面部识别系统是面部识别系统.面部地标检测 面部地标检测估计头部姿势的估计.多任务学习学习伪标签是一种伪标签.

更多相关视频

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.7K
Author Spotlight: Deciphering Electrical Networks Behind Complex Brain Activities and Disorders
05:49

Author Spotlight: Deciphering Electrical Networks Behind Complex Brain Activities and Disorders

Published on: November 1, 2024

772

相关实验视频

Last Updated: Jun 23, 2026

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

9.0K
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.7K
Author Spotlight: Deciphering Electrical Networks Behind Complex Brain Activities and Disorders
05:49

Author Spotlight: Deciphering Electrical Networks Behind Complex Brain Activities and Disorders

Published on: November 1, 2024

772

科学领域:

  • 计算机视觉和机器学习
  • 生物识别的人工智能

背景情况:

  • 面部分析模型在现实场景中经常失败,原因是学习各种人类特征和背景噪音的局限性.
  • 现有的面部标志检测和头部姿势估计的训练数据集经常是有限的和杂的,阻碍了概括.
  • 在面部分析模型在标准化测试中的表现与现实应用之间存在很大的差距.

研究的目的:

  • 为了弥合标准化和现实世界面部分析测试之间的性能差距.
  • 为了提高面部地标检测,头部姿势估计和面部识别的稳定性和准确性.
  • 开发一个框架,克服面部分析任务中各种培训数据的局限性.

主要方法:

  • 提出了一种伪标签技术,利用多样化的面部识别数据集来增强训练数据.
  • 开发了一个综合框架,使用互补的多任务学习来进行强大的特征提取.
  • 结合伪标签与多任务学习,以促进对姿势不变特征的学习,以改善面部识别.

主要成果:

  • 在AFLW2000-3D和BIWI数据集上实现了最先进的 (SOTA) 或接近SOTA的性能,用于面部地标检测和头部姿势估计.
  • 在IJB-C数据集上表现出具有竞争力的面部验证性能.
  • 即使在训练数据集中缺乏多样化的面部识别,也表现出稳定的表现,通过新的软,中和硬案例分类来验证.

结论:

  • 拟议的伪标签和多任务学习框架显著提高了面部分析在具有挑战性的现实环境中的性能.
  • 该方法有效地解决了培训数据集中缺乏数据多样性和噪音问题.
  • 综合方法带来了更强大的和可通用的面部分析系统.