Fusion of Convolutional Neural Networks and Random Forests for Accurate Potato Leaf Disease Detection

In order to classify potato leaf diseases, this work suggests a hybrid model that combines a Random Forest classifier for sample class classification with a CNN for image feature extraction. The dataset obtained from Kaggle includes 1885 samples of different potato leaf illnesses caused by fungi, viruses, pests, bacteria, Phytophthora, nematodes, and healthy leaves. Thorough preprocessing entails dividing the dataset into equal training, testing, and validation sets and downsizing photos to a standard dimension. The CNN model uses ReLU activation for normalization and softmax activation for classification. It has five layers with progressively larger filter sizes. For the final label classification, features taken from the CNN are fed into a Random Forest classifier consisting of 100 decision trees. The suggested model shows promise in terms of performance and accuracy in classifying diseases. The hybrid model suggested for classifying diseases affecting potatoes shows encouraging results; it has an overall accuracy rate of 77% and a precision range of 44% to 96% for various disease categories.