Large image/videos’ automatic annotation using bidirectional associative memories

In this paper, an approach of image/video´s automatic annotation considering semantics´ tolerance relation between pre-defined classes is described. The image/videos´ semantics is represented by a vector of associative values with the pre-defined classes. Moreover, the semantic gap induced in the process of automatically generating associative values is dealt with by using the constructed semantic tolerance relation model (STRM). On the other hand, the method how to generating associative values with pre-defined classes regarding different dimensions..is also depicted. Image/videos´ retrieval results based on the pre-defined classes shows the effectiveness of the proposed methods.