Research on image retrieval based on artificial intelligence technology

Image retrieval holds significant importance within the realm of computer vision. This paper introduces SCDNet, a novel network model, leveraging selected feature aggregation for enhanced image retrieval. Using Thangka images as an illustrative example, the proposed model aims to address issues such as inadequate semantic feature extraction and image feature distortions due to background or noise. SCDNet firstly uses VGG16 network to extract semantic features from different convolutional layers of Thangka images, then passes through feature screening module, feature aggregation module, and rearrangement module, thereby completing the retrieval. Our experimental validation on homemade Thangka image dataset and publicly available fine-grained image dataset, demonstrating the superiority of SCDNet compared with other retrieval algorithms.