iQIYI-VID
收藏资源简介:
iQIYI-VID是由爱奇艺公司创建的大规模视频数据集,专门用于多模态人物识别研究。该数据集包含60万个视频片段,涉及5000名名人,这些视频片段是从40万小时的在线视频中提取的,涵盖电影、综艺节目、电视剧和新闻广播等多种类型。所有视频片段都经过了严格的人工标注,标签错误率低于0.2%。iQIYI-VID数据集旨在推动多模态人物识别技术的发展,通过结合面部、头部、身体和音频等多种特征,提高人物识别的准确性。数据集的创建过程包括从大量视频中提取片段、自动过滤和人工标注等步骤,确保了数据集的质量和实用性。该数据集适用于评估和改进人物识别算法,特别是在复杂和多变的视频环境中。
iQIYI-VID is a large-scale video dataset created by iQIYI, specifically dedicated to multimodal person recognition research. This dataset contains 600,000 video clips involving 5,000 celebrities, which are extracted from 400,000 hours of online videos, covering various genres such as movies, variety shows, TV dramas, and news broadcasts. All video clips have undergone strict manual annotation, with a label error rate of less than 0.2%. The iQIYI-VID dataset aims to promote the development of multimodal person recognition technologies, by combining multiple features including facial, head, body, and audio information to improve the accuracy of person recognition. The dataset creation process includes steps such as extracting clips from massive videos, automatic filtering, and manual annotation, ensuring the quality and practicality of the dataset. This dataset is suitable for evaluating and improving person recognition algorithms, especially in complex and variable video environments.
- 1iQIYI-VID: A Large Dataset for Multi-modal Person Identification爱奇艺公司 · 2019年



