OMG-Emotion

该数据集名为OMG-Emotion，由大约一分钟长的YouTube视频组成，这些视频根据连续的情感行为进行了标注。总共收集了675个视频，总时长约为10小时。每个视频中的发声都带有两个连续的标签，分别代表情感唤起和情感倾向。这些情感表达受到个人特定特征的影响。该数据集规模为675个视频，总时长约10小时，其任务是进行情感识别。

The dataset named OMG-Emotion comprises approximately one-minute-long YouTube videos annotated according to continuous affective behaviors. A total of 675 videos have been collected, with an aggregate duration of roughly 10 hours. For each video, the utterances therein are paired with two sequential tags, which respectively correspond to emotion arousal and emotional valence. These emotional expressions are influenced by individual-specific traits. With a total of 675 videos spanning approximately 10 hours, the core task of this dataset is emotion recognition.

搜集汇总

背景与挑战

背景概述

OMG-Emotion数据集是一个专注于长期情感行为分类的挑战性语料库，旨在支持情感计算领域的研究，特别是在自然情感表达识别方面。该数据集为研究人员提供了一个评估和比较模型的平台，以推动更自然的情感描述模型的发展。