Design of Printed Fabric Image Retrieval System based on Deep Learning

Fabric image retrieval has great significance in fabric styling, procurement and management etc. in the textile industry. Given the time-consuming and inefficiency of manual retrieval, a simple yet effective retrieval system for rapid printed fabric images based on a modified AlexNet is proposed. Firstly, the pre-trained model AlexNet is fine-tuned to acquire a Convolutional Neural Network (CNN) model that is adaptable to printed fabrics. Then the model is used to extract features of the datasets. Subsequently, the similarities between query and database image features are calculated by Euclidean distance. Finally, the top k images and label information of the query image are returned to the clients. In this paper, the effectiveness of the proposed system is validated based on the printed fabric datasets. The results show that the performance of the proposed method is superior to other methods on printed fabric images in terms of retrieval.