FERV39k
收藏资源简介:
FERV39k是一个大规模多场景的人脸表情识别视频数据集,由复旦大学工程与应用技术研究院和计算机科学学院联合创建。该数据集包含38,935个视频片段,每个片段标注了7种基本表情之一,涵盖了日常生活中的4种场景,细分成22个具体场景。数据集的创建过程包括从互联网上自动获取候选视频片段,并通过精心设计的工作流程进行手动标注,确保标签的高质量。FERV39k数据集的应用领域广泛,旨在解决视频中人脸表情识别的实际问题,特别是在复杂和多变的场景中。
FERV39k is a large-scale multi-scenario video dataset for facial expression recognition, jointly developed by the Institute of Engineering and Applied Technology and the School of Computer Science at Fudan University. This dataset contains 38,935 video clips, each annotated with one of seven basic facial expressions, covering four general daily scenarios and subdivided into 22 specific scenarios. The dataset construction process involves automatically acquiring candidate video clips from the Internet, followed by manual annotation via a meticulously designed workflow to ensure high annotation quality. The FERV39k dataset has wide-ranging application scenarios, aiming to address practical challenges in facial expression recognition from videos, especially in complex and variable scenarios.

- 1FERV39k: A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos复旦大学工程与应用技术研究院 · 2022年



