A Hybrid Plant Disease Detection Algorithm Using Residual MBi-LSTM With CNN Model

In India, various plant diseases affect agricultural productivity. For this reason, crop losses occur every year. On-time, the accurate detection of all diseases is essential to ensure healthy plants and can lead to improved yields. Traditionally, we needed the expertise of an agricultural specialist. However, in recent years, numerous deep-learning methods have been introduced, promising to automate the diagnosis of plant diseases using the images of infected plants. Despite these achievements, many existing models fail to function effectively when data are altered according to time and place. To address this problem, we propose a model that combines VGG16 with a multilayer bidirectional long short-term memory (MBi-LSTM) network. The VGG16 component captures spatial hierarchies and extracts features in the images. The MBi-LSTM layers learn temporal relationships across image sequences. By integrating both spatial and temporal information, our hybrid approach achieves a deeper understanding of visual patterns as compared to models that rely solely on spatial features. We use two datasets (PlantVillage and real world) for training and testing our proposed model of labeled plant disease images. Quantitative results demonstrate that, across all evaluation metrics—accuracy, precision, recall, andF1-score—the VGG16 + MBi-LSTM model achieved the highest performance. The classification accuracy achieved by the model on the PlantVillage dataset is 98.9% and on the real-world dataset is 96.6%, showcasing its effectiveness for real-time disease detection. This method provides a reliable solution for disease prediction, enabling farmers to take preventive measures at an early stage of the crop’s development.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.