A web-based digital library for finding similar image using neural network

Abstract

In the advancing digital era, the vast amount of information on the internet has made it increasingly challenging to find relevant and similar resources, including books and documents. This has led to the development of digital libraries that utilize supportive technologies for information retrieval and recommendation, encompassing both text and image-based content. This project focuses on creating a web-based digital library that employs neural network techniques to facilitate the retrieval of book-related information using images. The limitations of physical libraries, such as disrupted research activities and time consuming processes of searching for available books, especially considering the large volume and diverse subjects, can be addressed through a digital library. Additionally, while keyword based (text, field, title, author, etc) searching is the primary method in digital libraries, incorporating image-based search, as effectively demonstrated by e-commerce websites, can enhance the user experience and provide more accurate search results. To address these issues, this project builds a digital library website with basic functionalities for users, including book searching, viewing book information, and selecting reservation dates. Furthermore, the addition of image-based search functionality using a machine learning model based on neural networks will be implemented. Through the research and development process, a successful web-based digital library has been built, providing book information to users and allowing them to check the availability of books. The project also includes a dashboard for administrators to manage the data of the digital library. Moreover, training a machine learning model based on neural networks to extract image features and perform similar image retrieval has demonstrated its compatibility with book cover images and accurate search results. To enhance the performance of the project, the utilization of optimization techniques using neural networks can improve the model's output. Additionally, a model for cropping the main content of input images can contribute to improving the efficiency of the information retrieval function. Regarding the digital library website, adding more diverse data to align with real-world scenarios and optimizing the interaction between administrators and the system's dashboard is crucial.