BriefGPT.xyz
Sep, 2024
数据集中的错误标签对人类姿态估计的影响
The Influence of Faulty Labels in Data Sets on Human Pose Estimation
HTML
PDF
Arnold Schwarz, Levente Hernadi, Felix Bießmann, Kristian Hildebrand
TL;DR
本研究探讨了训练数据质量对人类姿态估计模型性能的影响，特别是错误标签如何对学习和性能指标产生负面影响。通过对流行数据集进行深入分析，揭示了标签不准确的程度和性质，从而提出考虑错误标签影响的方法，以促进更鲁棒和准确的人类姿态估计模型的发展。
Abstract
In this study we provide empirical evidence demonstrating that the quality of training data impacts
Model Performance
in
Human Pose Estimation
(HPE). Inaccurate labels in widely used data sets, ranging from minor
→