Tomato Plant Leaf Diseases Classification using Enhanced LSTM and BiLSTM

Considered the economic backbone of many nations, tomatoes are among the most valuable crops in the world. However, these harvests are still vulnerable to a number of diseases that can diminish and even wipe out the production of healthy crops; thus, early and accurate disease detection is essential. In recent years, several research studies have used deep learning (DL) models for autonomous tomato leaf disease diagnosis. Unfortunately, a lot of these techniques rely on a single DL architecture, which increases the classification difficulty because it requires a lot of processing power to update these hyper-parameters. Furthermore, they retrieved huge dimensions from these networks, which made categorization much more difficult. As a result, we used enhanced LSTM and BiLSTM models to classify the tomato leaf diseases. In this work, we used 31,793 tomato leaf images consisting of ten disease classes with healthy images for the classification. We used Jupyter Notebook to simulate the classification-enhanced LSTM and BiLSTM. The proposed model utilized CLAHE for preprocessing the images, VGG19 was used as a feature extractor, and LSTM and BiLSTM were used as classifiers. VGG19 provides an effective and robust feature extraction to classify the leaf diseases. The proposed model achieved an accuracy of 97.33% and 98.49% for VGG19+LSTM and VGG19+BiLSTM for the classification of the tomato leaf diseases. This automated classification provides efficient and robust classification of leaf diseases in early stages, improves the crop quality, and increases the outcomes.