基于层次聚类的图文检索模型研究
Research on Image and Text Retrieval Model Utilizing Hierarchical Clustering
孙健玮 1刘玉龙1
作者信息
- 1. 中国电子科技集团公司第15研究所,北京 100083
- 折叠
摘要
图文检索在工业中的用途和作用是多方面的,可以帮助提高研发和生产效率,促进科技创新,提高产品的质量和竞争力;目前,图文检索模型的重点是提高检索的精度;随着技术和数据的快速发展,深度学习和大模型技术的不断应用,图文检索的速度问题逐渐凸显,为解决当前图文检索速度受限、计算量大的问题,提出了一种基于层次聚类的图文检索模型;该方法选择了检索效果明显的跨模态哈希方法,并运用深度聚类算法对待检索的数据进行分类,从而缩小检索范围,提高了检索速度;实验结果表明,基于层次聚类的图文检索模型在保持检索精度的同时,显著提高了检索速度,使得工程人员能够更快地获取到满意的检索结果.
Abstract
The application and impact of image-text retrieval in industry are multifaceted,as it can help improve research and de-velopment efficiency,promote technological innovation,and enhance product quality and competitiveness.Currently,the emphasis of image-text retrieval models is on improving retrieval accuracy.With the rapid development of technology and data,the continuous ap-plication of deep learning and large-scale model techniques has gradually highlighted the issue of retrieval speed in image-text retrieval.To address the current limitations in retrieval speed and high computational requirements,a hierarchical clustering-based image-text retrieval model has been proposed.This method adopts a cross-modal hashing approach with evident retrieval effectiveness and applies deep clustering algorithms to classify the data to be retrieved,thereby reducing the retrieval scope and improving retrieval speed.Ex-perimental results indicate that the hierarchical clustering-based image-text retrieval model significantly enhances retrieval speed while maintaining retrieval accuracy,enabling engineering personnel to obtain satisfactory retrieval results more quickly.
关键词
图文检索/跨模态哈希方法/深度学习/深度聚类算法/信息检索Key words
image-text retrieval/cross-modal hashing/deep learning/deep clustering/information retrieval引用本文复制引用
出版年
2024
NETL
NSTL
万方数据