基于视觉内容与语义相关的图像标注模型

An Image Annotation and Refinement Model Based on Visual Content and Semantic Correlation

针对当前标注系统的不足,设计了一种高效的标注模型,其标注步骤包括标注和标注改善,标注算法采用加权的正反例标志向量法,标注改善采用NGD方法。实验表明,标注效率远优于经典的标注模型,标注质量优于大多数标注模型。

The efficiency and qulaity of image annotation system determine the ability to manage images in the fields of computer vision and image retrieval. To overcome the drawback of current annotation system, an efficient annotation system is designed , including annotation and refine- ment stages by weighted positive and negative symbol vector method and NGD method respec- tively. The experiments demonstrate our proposed system perfomance, whose efficiency outper- forms classieial image annotation models and qulity outperforms most current image annotation models.