Plant Disease Classification Using Transfer Learning with ResNet Architecture

Abstract

This paper presents a neural network-based approach for classifying plant leaf diseases using deep learning. Initially, a custom Convolutional Neural Network (CNN) was developed, followed by experiments with deeper pretrained architectures such as VGG16 and ResNet50. Among them, ResNet50 achieved the highest classification accuracy, demonstrating superior learning capability and robustness. The model was trained on a publicly available plant disease dataset containing 38 classes, enhanced through data augmentation techniques. Transfer learning and fine-tuning were employed to improve model efficiency and accuracy. The primary objective of this work is to compare deep learning architectures and identify the most effective model for real-time plant disease diagnosis. Experimental results confirm that the ResNet50 model outperforms the others in both training convergence and predictive accuracy.

Plant phenotyping relevance

植物葉画像から病害状態を推定する深層学習分類法を開発・比較しており、植物表現型（病害状態）の取得・推定が中心課題であるため含める。

Code and data availability

保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。