WCSE

Using YoloV8 Book Defect Detection Model with MIRNet LLI

Abstract— Books serve as fundamental repositories of knowledge. In an age where digitization of texts is rampant, the tangible versions of these tomes are at risk of degradation due to environmental factors, mishandling, and the natural decay of materials. The integrity of a book, both in terms of its legibility and visual appeal, can be compromised by various forms of damage. It is crucial for institutions like museums and libraries to preserve these concrete sources of knowledge. Traditional methods for identifying damages in books have relied on manual checks and rudimentary image processing. However, the inefficiencies of these methods have highlighted the urgent need for more sophisticated, automated systems for the detection of book defects. This paper introduces a novel approach that utilizes Roboflow and YOLOv8, an advanced object detection model, in conjunction with MIRNet for Low-Light Image (LLI) enhancement to detect and classify various book defects. Our proposed system aims to significantly reduce human effort and increase the accuracy of defect detection. Through extensive experiments and evaluations, we demonstrate the effectiveness of our model in identifying a range of defects such as torn pages, water damage, and broken bindings under various lighting conditions. The integration of MIRNet LLI ensures that the defects are accurately detected even in low-light images, which is common in the storage environments of many historical books. The outcomes suggest that our model not only streamlines the preservation process but also provides a scalable solution for libraries and archives worldwide.

Index Terms— Computing methodologies, Artificial intelligence, Computer vision, Object detection, Computer graphics, Image processing

John Paul Tomas
Mapúa University, PHILIPPINES
Katrice Asher Albano
Mapúa University, PHILIPPINES
Jill Samantha Alday
Mapúa University, PHILIPPINES
Matthew Raphael Corbe
Mapúa University, PHILIPPINES
Gian Nicole Pangan
Mapúa University, PHILIPPINES

[Download]

Cite: John Paul Tomas, Katrice Asher Albano, Jill Samantha Alday, Matthew Raphael Corbe, Gian Nicole Pangan, "Using YoloV8 Book Defect Detection Model with MIRNet LLI," 2024 The 14th International Workshop on Computer Science and Engineering (WCSE 2024), pp. 27-35, Phuket Island, Thailand, June 19-21, 2024.