Detection of Rice and Corn Plant Leaf Disease Using Invariants of Deep Learning Models and Edge Perspective

Abstract

Rice and corn hold significant importance due to their daily consumption worldwide. Naked-eye observations are not accurate. Therefore, we need an autonomous system that can accurately detect and classify diseases in both plants. We trained and validated publicly available datasets in three deep convolutional neural network (DCNN)-based deep learning models using different learning rates and found that the lowest learning rate was the most effective in achieving the highest accuracy. We added a new dense layer to the known DCNN-based deep learning models and achieved improved accuracy. The best results were observed when our invariants of the InceptionV3, ResNet152, and MobileNetV2 deep learning models were used on corn plant leaves (98.09%, 98.51%, and 89.73%, respectively). These models also performed well on rice plant leaves (98.51%, 93.59%, and 98.57%, respectively). Because InceptionV3 performed well for both plants, we implemented it in NVIDIA Jetson Nano as an end device for the detection and classification of diseases from both plant leaves. Received: 4 January 2025 | Revised: 26 May 2025 | Accepted: 13 June 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 Kaggle at https://www.kaggle.com/datasets/vbookshelf/rice-leaf-diseases and https://www.kaggle.com/datasets/smaranjitghose/corn-or-maize-leaf-disease-dataset. Author Contribution Statement Zubair Saeed: Conceptualization, Methodology, Software, Validation, Resources, Data curation, Writing – original draft, Project administration. Uzma Nawaz: Validation, Formal analysis, Writing – review & editing. Ali Raza: Conceptualization, Validation, Formal analysis, Writing – review & editing, Visualization. Kamran Javed: Validation, Formal analysis, Investigation, Writing – review & editing, Visualization, Supervision.

Plant phenotyping relevance

植物葉の病徴を画像から検出・分類する深層学習手法を開発・検証し、エッジデバイスにも実装しており、植物状態の取得・推定が研究の中心です。

Code and data availability

The paper's Data Availability Statement explicitly names two public Kaggle image datasets (rice leaf diseases and corn/maize leaf disease) that were used directly as the phenotyping inputs for this study's disease-classification experiments. No author code, models, or checkpoints are reported as publicly available.