Rice leaf disease classification using a fusion vision approach.

Abstract

Rice serves as a fundamental staple for a significant portion of the global population, playing an essential role in ensuring food security worldwide. However, the continuous threat of various diseases risks both yield and quality. Detecting these diseases at an early stage is very important for effective management of these risks. This research introduces a novel approach for rice disease detection using the fusion vision boosted classifier (FVBC), integrating VGG19 for feature extraction and LightGBM for classification. The meticulously curated dataset comprises 2627 rice leaf images, categorized into training, validation, and test sets for robust model evaluation. The FVBC model achieves impressive accuracies of 97.78% on the training set, 97.5% on the validation set, and 97.6% on the test set, demonstrating its efficacy in disease detection. The model's performance compared with other classifiers, including Softmax, highlights its superiority. Hyperparameter tuning, such as learning rate and tree depth for LightGBM, was crucial for optimizing model performance. The proposed FVBC model offers a non-invasive, scalable solution for early disease detection, empowering farmers to implement timely interventions and enhance agricultural productivity.

Plant phenotyping relevance

イネ葉画像から病害状態を推定する分類手法を開発・評価しており、植物病害表現型の取得が研究の中心である。

Code and data availability

The paper's rice leaf disease image dataset (2627 images, six classes) is publicly available on Kaggle and is the paper-specific input used for the FVBC phenotyping/classification analysis. However, the Kaggle dataset URL (reference 42) does not appear in the allowed_urls list, and no author analysis code, trained FVBC

No evidence-backed public reproduction asset is currently recorded.