Handwritten Cyclic Compound Classification with EfficientNetB0 and Explainable AI Methods

DOI

10.2991/978-94-6239-664-7_48How to use a DOI?

Keywords

Handwritten Chemical Structure Recognition; Explainable Artificial Intelligence; Grad-CAM; SHAP; Cyclic Compound

Abstract

Handwritten chemical structures are commonly used in chemistry education and research, but recognizing these drawings automatically remains a challenging task. Variations in writing styles, line thickness, and incomplete patterns make accurate classification difficult for computer systems. This study presents a deep learning-based framework for the automatic identification of handwritten cyclic compounds. We created a dataset of handwritten chemical compounds that contains 599 images. Then, data augmentation was applied, resulting in a total of 15945 images across 15 chemical compounds in this study. Five pre-trained Convolutional Neural Network (CNN) architectures, such as VGG16, VGG19, ResNet101, MobileNetV2, and EfficientNetB0, were fine-tuned and used in this research. Among them, EfficientNetB0 achieved the best performance with an accuracy of 97.87%, showing strong generalization and minimal overfitting. To enhance interpretability, Explainable AI (XAI) techniques such as SHAP, Grad-CAM, and Saliency Maps were applied. Each of the XAI highlights the molecular regions that most influenced the model’s decisions. The visualization results confirmed that the model focused on key chemical structures such as ring systems and functional groups. Overall, this research provides an interpretable and accurate approach for recognizing handwritten cyclic compounds.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Md. Ismiel Hossen Abir
AU  - Sanjana Islam Kasfia
AU  - Abir Mahmud Shahariar
PY  - 2026
DA  - 2026/06/08
TI  - Handwritten Cyclic Compound Classification with EfficientNetB0 and Explainable AI Methods
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 688
EP  - 700
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_48
DO  - 10.2991/978-94-6239-664-7_48
ID  - Abir2026
ER  -