CNN-LSTM Learning Approach for Classification of Foliar Disease of Apple

Deep learning techniques haven been the efficient artificial intelligence techniques that is being used for developing models for autonomous feature extraction, classification and segmentation of the apple leaves diseases sand other plant diseases. However, most deep learning systems, such as Convolutional neural networks (CNN) and Long Short-term Memory (LSTM), necessitate a huge quantity of training data and are plagued by issues such as inflating gradients, overfitting, and class imbalance, among others. In this work, CNN-LSTM deep learning algorithm was proposed to address the challenges facing CNN and LSTM respectively. This work used the CNN-LSTM deep learning algorithm to develop a learning model for classification of foliar disease of apple leaves. In order to create learning classification models based on accuracy, specificity, sensitivity, and AUC standard performance evaluation techniques, both CNN and LSTM algorithms were employed. CNN-LSTM happened to be the best model in terms of accuracy, specificity, sensitivity, and AUC standard performance evaluation methodologies, with 98.00%,95.00%, 96.00parcent, and 94.00% respectively. Hence, the model has shown a significant improvement as compared to other models for being able to correctly classified apple leaves into affected leaves with disease/s and non-affected leaves with disease/s with 98.00% accuracy. Likewise, for the ability to classify the apple leaves affected with disease/s, correctly CNN-LSTM also outperformed other models with 95.00parcent. Moreover, CNN-LSTM based model also has shown a significant improvement as compared other models, for being able to correctly classified apple leaves that were not affected with disease/s with 96.00%.