大型自然主义数据集

本研究使用了一个包含约55,000个视频的大型自然主义数据集，这些视频记录了来自90多个国家的参与者在自然环境下的面部表情。数据集中的视频经过专业训练的FACS编码员标注了不同的动作单元（AU）。数据集的创建旨在解决面部表情识别中的自然性和多样性问题，通过提供大量自然环境下的面部表情数据，支持深度学习模型在动作单元检测任务中的训练和评估。

This study employs a large-scale naturalistic dataset containing approximately 55,000 videos, which capture facial expressions of participants from over 90 countries in naturalistic environments. All videos in the dataset were annotated with distinct Action Units (AUs) by professionally trained FACS coders. This dataset was developed to address the challenges of naturalness and diversity in facial expression recognition, and by providing abundant naturalistic facial expression data, it supports the training and evaluation of deep learning models for Action Unit detection tasks.