Automated Diagnosis of Marigold Leaf Diseases using a Hybrid CNN-SVM Model

For healthy yields of crops and to maintain food security, it is essential to identify and classify plant diseases. In this research, the aimed to create a convolutional neural network-based deep learning algorithm for accurately classifying various marigold leaf diseases. 2000 images of marigold leaves were included in the dataset, with 200 images for each of the ten groups. (nine diseases and one healthy leaf). Six layers of convolution, six maximum-pooling layers, and two fully linked layers with regularisation and ReLU function activation made up the suggested CNN model. The efficiency of the model was optimized through hyperparameter tuning. Using the Keras framework, the model was trained and evaluated on the preprocessed dataset, and its performance was assessed using metrics like precision, recall, F1-score, as well as accuracy. According to the findings, the developed model classified the various classes of marigold leaves with an overall accuracy of 92%, with precision, recall, as well as F1-score metrics varying from 50% to 71.79% depending on the class. The confusion matrix gave a thorough summary of how well the model did at identifying each of the various marigold leaf classes. The created model can be applied to the early discovery of diseases affecting marigold leaves, aiding in both disease containment and yield protection. This research also emphasizes how well deep learning models work at accurately identifying and categorizing various plant diseases.