多模态情感数据集

多模态情感数据集涵盖视觉、听觉两种模态，提供更全面的情感信息。每种情感类别包含2000条数据，单一数据长度超过3s，总量达10000条，保证模型训练的充分性和泛化能力，并且数据总量持续扩充中。该数据集涵盖了开心、悲伤、生气、害怕和中性这五种基本情感类别，每个表达情绪的视频帧中仅包含一个人脸。数据集包含不同年龄段、性别和种族的人员，以确保情感的多样性和普遍性。该数据集涵盖了开心、悲伤、生气、害怕和中性这五种基本情感类别，每个表达情绪的视频帧中仅包含一个人脸。数据集包含不同年龄段、性别和种族的人员，以确保情感的多样性和普遍性。多模态融合的情感分析提供了高质量的融合数据，可以用于多模态融合大模型（如多模态BERT、多模态Transformer等）的训练，将视觉、听觉进行融合分析，获取更全面的情感信息，推动多模态大模型情感识别技术的进步。

This multimodal sentiment dataset covers two modalities, visual and auditory, to deliver more comprehensive affective information. Each emotion category contains 2000 samples, with each individual sample lasting over 3 seconds. The total number of samples reaches 10,000, which ensures the sufficiency and generalization capability of model training, and the dataset scale is continuously expanding. This dataset includes five basic emotion categories, namely happiness, sadness, anger, fear and neutrality. Each video frame expressing emotions contains only one human face. The dataset covers individuals of different ages, genders and ethnicities to ensure the diversity and universality of emotional expressions. This dataset includes five basic emotion categories, namely happiness, sadness, anger, fear and neutrality. Each video frame expressing emotions contains only one human face. The dataset covers individuals of different ages, genders and ethnicities to ensure the diversity and universality of emotional expressions. The multimodal fused sentiment analysis provides high-quality fused data, which can be employed for training multimodal large models such as multimodal BERT, multimodal Transformer, etc. By fusing visual and auditory information to conduct comprehensive affective analysis, it can promote the advancement of emotion recognition technologies for multimodal large models.