【摘要】 笔者提出了一种监督式层次狄利克雷过程模型(MM_HDP)。该模型利用层次狄利克雷过程模型发现文本与图像间的关联主题,解决异质多模态数据存在的语义鸿沟问题,具备实现图像自动标注与分类的能力。为了验证模型的有效性,在LabelMe和UIUC数据集上进行了测试。实验结果表明,相对于其他传统模型,该模型表现出了良好的效果。更多还原

【Abstract】 A supervised hierarchical Dirichlet process model (mm) is proposed_ HDP).The model uses hierarchical Dirichlet process model to find the related topics between text and image,and solves the semantic gap problem of heterogeneous multimodal data,and has the ability to realize automatic image annotation and classification.In order to verify the validity of the model,it is tested on labelme and UIUC datasets.The experimental results show that the model has good performance compared with other traditional models.更多还原