BlanketGen-3DPW

BlanketGen-3DPW是由生物医学工程研究中心创建的一个增强型数据集，旨在解决临床环境中人体姿态估计的挑战。该数据集通过BlanketGen管道，对原始的3DPW数据集进行了增强，添加了合成毯子遮挡，以模拟真实世界中的遮挡情况。数据集包含1037个视频，总计96399帧，主要用于训练和测试深度学习模型在有遮挡情况下的性能。创建过程涉及使用Blender软件进行3D建模和渲染，确保毯子遮挡的真实性和复杂性。该数据集特别适用于研究在床场景中的人体姿态估计，如癫痫监测和睡眠分析，以提高模型在实际应用中的准确性和鲁棒性。

BlanketGen-3DPW is an enhanced dataset developed by the Biomedical Engineering Research Center, which aims to address the challenges of human pose estimation in clinical settings. This dataset enhances the original 3DPW dataset via the BlanketGen pipeline by adding synthetic blanket occlusions to simulate real-world occlusion scenarios. It contains 1037 videos with a total of 96,399 frames, and is primarily used for training and testing the performance of deep learning models under occlusion conditions. The creation process utilized Blender software for 3D modeling and rendering, ensuring the authenticity and complexity of the simulated blanket occlusions. This dataset is particularly suitable for research on human pose estimation in bed scenes, such as epilepsy monitoring and sleep analysis, to improve the accuracy and robustness of models in practical applications.