VidHarm

VidHarm是一个由瑞典媒体委员会和林雪平大学计算机视觉实验室合作创建的开放数据集，包含3589个来自电影预告片的10秒视频片段，由专业电影分类师进行标注。数据集旨在解决自动识别视频中有害内容的问题，特别是在儿童保护方面。数据集的创建过程涉及从电影预告片中提取片段，并由专业人员根据瑞典电影分类系统进行标注。该数据集适用于视频识别模型的训练和评估，特别是在年龄评级分类方面。

VidHarm is an open dataset co-created by the Swedish Media Council and the Computer Vision Laboratory of Linköping University. It contains 3,589 10-second video clips sourced from movie trailers, which were annotated by professional film classifiers. This dataset aims to address the challenge of automatically identifying harmful content in videos, with a particular focus on child protection. The creation of the dataset involved extracting clips from movie trailers, and having professionals conduct annotations in accordance with the Swedish film classification system. It is suitable for the training and evaluation of video recognition models, especially for age rating classification tasks.