Image Classification of Graphene Oxide Thin Films’ Sheet Resistance using a Convolution Neural Network

• Published in: 2024 IEEE 7th International Conference on Electrical, Electronics, and System Engineering: Dissemination and Advancement of Engineering Education using Artificial Intelligence, ICEESE 2024
• Main Author: Yusri M.W.B.; Masrie M.; Badaruddin S.A.M.; Burham N.; Janin Z.; Saad H.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85217396453&doi=10.1109%2fICEESE62315.2024.10828545&partnerID=40&md5=f208c37a81cd53bc6fbb06b3ce32aa28

This study focuses on developing a CNN model, VGG-16, to classify microscopy images of graphene oxide thin films produced by two machines; Atomizer 2 and Atomizer 3 based on the sheet resistance values. The methodology begins with preparing microscopic images of graphene oxide thin films dataset. The dataset undergoes preprocessing to enhance image quality. It is then divided into training (80%) and testing (20%) sets. Data augmentation techniques were applied to improve the model's generalization capabilities. The core of this research involves constructing a CNN model using the VGG-16 architecture, which is trained on the preprocessed dataset. Training and validation results are obtained to assess the model's performance. Subsequently, a separate test model evaluates the accuracy of the image classification process. The results indicate an accuracy of 76.7% for images from Atomizer 2 and 92.37% for images from Atomizer 3, demonstrating the effectiveness of the developed AI program in classifying graphene oxide thin films microscopic images based on sheet resistance values. © 2024 IEEE.