caskcsg/Libra-Emo

Libra-Emo是一个大规模的多模态细粒度数据集，用于负面情绪检测。它包括视频片段、字幕、情绪标签和相应的解释。数据集是单语言的（英语），并遵循CC BY 4.0许可证发布。数据集的大小在10K到100K样本之间，分为训练集和测试集。特征包括视频、字幕、标签和解释。README提供了详细的下载、解压和使用数据集的说明，以及用于预测和评估的脚本。该数据集专为视频分类任务设计，特别是用于识别负面情绪。

Libra-Emo is a large-scale multimodal fine-grained dataset for negative emotion detection, including video clips, subtitles, emotion labels, and corresponding explanations. The dataset is monolingual (English) and distributed under the CC BY 4.0 license. The size of the dataset is between 10K and 100K samples, split into training and testing sets. Features include video, subtitle, label, and explanation. Detailed instructions for downloading, extracting, and using the dataset are provided, along with scripts for prediction and evaluation. The dataset is designed for video classification tasks, particularly for identifying negative emotions.