Klasifikasi Tingkat Keparahan Penyakit Leaf Blast pada Tanaman Padi menggunakan EfficientNetB0 menggunakan Optimasi CLAHE

Abstract

This study assessed the effectiveness of the EfficientNetB0 model in predicting the
severity of leaf blast disease in rice plants caused by the fungus Pyricularia oryzae. Using
a dataset of 300 photos classified into three severity levels (healthy-light, moderate, and
severe), the model obtained a validation accuracy of about 90%. However, the model
struggled to identify disease groups with “moderate” and “severe” severity, indicating a
bias in categorization. Training methods included data augmentation and fine-tuning
strategies to improve model performance. Evaluation findings showed that the model
was more accurate in identifying the “mild-healthy” class than the other classes. This
study emphasizes the relevance of AI-based solutions for plant disease detection and
provides suggestions for future development to improve classification accuracy.