OMG-Emotion (One-Minute Gradual-Emotional Behavior)

一分钟渐进情绪行为数据集 (OMG-Emotion) 数据集由长度约为一分钟的 Youtube 视频组成，并在考虑到连续情绪行为的情况下进行了注释。这些视频是使用爬虫技术选择的，该技术使用基于长期情感行为的特定关键字，如“独白”、“试镜”、“对话”和“情感场景”。它包含 567 个平均时长为 1 分钟的情感视频，这些视频来自各种 Youtube 频道。视频根据话语被分成几个片段，每个话语都由至少五个独立的受试者使用 Amazon Mechanical Turk 工具进行注释。

The One-Minute Progressive Emotional Behavior Dataset (OMG-Emotion) consists of approximately one-minute-long YouTube videos, annotated while accounting for continuous emotional behaviors. These videos were selected via web crawling technology that uses specific keywords based on long-term emotional behaviors, such as "monologue", "audition", "dialogue", and "emotional scene". It contains 567 emotional videos with an average duration of one minute, sourced from various YouTube channels. The videos are split into segments based on utterances, and each utterance was annotated by at least five independent annotators using the Amazon Mechanical Turk tool.

搜集汇总

背景与挑战

背景概述

OMG-Emotion数据集是一个专注于连续情绪行为分析的数据集，包含567个平均时长为1分钟的YouTube视频，这些视频通过特定关键词爬取，并被分割成片段，由至少五位独立受试者进行标注。该数据集由汉堡大学于2018年发布，适用于情感分析和视频分类任务。