Automated Evaluation of Smartphone Screen Damage: A CNN-Based Image Recognition System

Ogechi Chiama; Swathi Ganesan

doi:10.47852/bonviewAIA52024309

Authors

Ogechi Chiama Computer Science and Data Science Department, York St John University, UK
Swathi Ganesan Computer Science and Data Science Department, York St John University, UK https://orcid.org/0000-0002-6278-2090

DOI:

https://doi.org/10.47852/bonviewAIA52024309

Keywords:

convolutional neural network (CNN), screen damage classification, smartphone screens, machine learning, deep learning, mobile application, defect detection

Abstract

The automation of defect detection in smartphone screens is important in ensuring fairness in the refurbished smartphone market. Conventional methods for evaluating screen damage rely on manual assessments, which are subjective, unreliable, and susceptible to human error. This paper presents a deep-learning-based approach using a convolutional neural network (CNN) to classify smartphone screens as cracked or uncracked. The CNN model was trained using a custom dataset of smartphone images, with an accuracy rate of 92.0% in classifying screen damage. CNN outperformed conventional machine learning methods in terms of feature extraction, resulting in higher precision in defect detection. However, the model encountered difficulty in identifying subtle crack patterns, fluctuations in light conditions, and overfitting resulting from the dataset’s limited diversity. Future work will focus on expanding the dataset, refining methods for data augmentation, and exploring alternative algorithms such as a transformer-based or hybrid model to improve model quality. This study aims to facilitate the standardization of defect assessment in the refurbished smartphone industry by automating screen damage detection, thereby increasing consumer confidence and boosting resale market efficiency.

Received: 9 September 2024 | Revised: 16 April 2024 | Accepted: 22 May 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in RaYDaR at https://yorksj.figshare.com/.

Author Contribution Statement

Ogechi Chiama: Conceptualization, Methodology, Data curation, Writing – original draft, Writing – review & editing, Visualization. Swathi Ganesan: Supervision.