Optimizing rice plant disease detection with crossover boosted artificial hummingbird algorithm based AX-RetinaNet.

Abstract

Rice is the most important cereal food crop in the world, and half of the world's population uses rice as a staple food for its energy source. The yield production qualities and quantities are affected by biotic and abiotic factors namely viruses, soil fertility, bacteria, pests, and temperature. Rice plant disease is the most crucial factor behind communal, economic, and agricultural losses in the agricultural field. Farmers detect and identify diseases through the naked eye, which takes more time and resources, leading to crop loss and unhealthy farming. To overcome these issues, this paper presents a novel rice plant disease detection approach named the crossover boosted artificial hummingbird algorithm based AX-RetinaNet (CAHA-AXRNet) approach. This current research paper mainly concentrates on the effectiveness of rice plant disease detection and classification. The hyperparameters of the AX-RetinaNet model are optimized through the CAHA optimization model. In this paper, three types of disease detection datasets namely rice plant dataset, rice leaf dataset, and rice disease dataset are included to classify rice plants as healthy or unhealthy. The most essential performance metrics are precision, F1-score, accuracy, specificity, and recall, employed to validate the effectiveness of disease detection. The proposed CAHA-AXRNet approach demonstrates its effectiveness compared to other existing rice plant disease detection methods and achieved an accuracy rate of 98.1%.

Plant phenotyping relevance

イネ葉・植物の病害状態を画像データから検出・分類する手法を開発し、複数データセットと性能指標で評価しており、植物フェノタイピング手法が研究の中心である。

Code and data availability

The paper uses three public rice disease image datasets (Kaggle rice leaf, GitHub rice disease, Kaggle rice plant) as its phenotyping inputs; no author code or model release is stated (data availability is on-request only).