基于社会标签的图像情感自动分类标注研究

Automatic Image Emotion Categorization and Annotation Based on Social Tags

以现存的大量图像社会标签为基础,设计基于社会标签的图像情感自动分类标注模型,提出图像社会标签对情感的贡献度、区分度以及标签在情感维度上的权重3个因子及计算方法,实现图像5种基本情感语义的分类标注。在实际社会网络数据集上进行验证实验,结果表明建立的模型方法在自动标注图像情感语义方面具有较高的准确性和应用性。同时还发现,实际数据集中情感分布差异与复合情感对图像自动标注具有影响。

This paper, based on the existing massive social tagged images, designs an automatic image emotion categorization and annotation model based on these social tags, introduces three factors and their calculation methods, including weight to emotion, discrimination to emotion and contribution to emotion of a social tag, which can categorize images into five basic emotions and annotate each image with the corresponding emotion. Finally experiments are designed, and detailed data analysis is presented on a real social tagging website.