基于双重注意力机制的书画描述文本自动生成模型  

Model of Automatic Generation of Description Texts for Calligraphy and Painting Based on Dual Attention Mechanism

作　　者：冯萍[1] 黄彪 FENG Ping;HUANG Biao(College of Computer Science and Technology,Changchun University,Changchun 130022,China;College of Cyber Security,Changchun University,Changchun 130022,China)

机构地区：[1]长春大学计算机科学技术学院,长春130022 [2]长春大学网络安全学院,长春130022

出　　处：《长春大学学报》2023年第10期1-6,共6页Journal of Changchun University

基　　金：吉林省教育厅项目(JGJX2023D397)。

摘　　要：提出了一种基于双重注意力机制的书画描述文本自动生成模型。结合了编码器和解码器,并在其中引入注意力机制,以提高模型对书画作品的关注度和表达能力。通过在编码器中引入属性注意力机制,能够捕捉到书画作品的高级特征和视觉信息。而采用基于Transformer的解码器可以有效地生成准确、流畅的描述文本。验证了模型的有效性,并与现有的主流方法进行了比较。结果表明,模型在生成书画描述文本方面取得了显著的改进,并具有广泛的应用潜力。An automatic generation model of description texts for calligraphy and painting based on dual attention mechanism is proposed,in which the encoder and decoder are combined,and the attention mechanism is introduced,in order to improve the attention and expression ability of the model on calligraphy and painting works.The advanced features and visual information of calligraphy and painting works can be captured by introducing the attribute attention mechanism into the encoder,and the accurate and vivid description texts can be effectively generated by using Transformer-based decoder.The effectiveness of the model is verified by comparing with the existing mainstream methods.The results show that the model has made significant improvements in the generation of description texts for calligraphy and painting,having broad application potential.

关 键 词：注意力机制 TRANSFORMER 图像描述生成 视觉特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]