Deep Learning based method to predict Plant Diseases: A case study with Rice Plant Disease Classification

Bangladesh, a nation that largely depends on Rice for both its economy and dietary needs, is the target of the project, which intends to establish an easy and effective approach for recognizing rice plant diseases there. The research made use of an original dataset consisting of 5932 pictures of four various forms of Rice plant diseases. In order to extract features, we leveraged Convolutional Neural Networks (CNN) with Transfer Learning techniques and pre-trained models like VGG16, Inception-V3, VGG19, and ResNet-50. To increase the training and test accuracy, augmentation methods including flipping, zooming, and resizing the dataset were also performed. The VGG-19 model exceeded all other models with a test accuracy of 98%. Additionally, traditional machine learning models, including KNN, SVM, AdaBoost, Decision Tree, and Random Forest, were employed, with the Random Forest classifier achieving the highest accuracy of 96.6%. Gabor and Sobel filters were utilized for feature extraction before feeding the datasets into the machine-learning models. The suggested model performed better than the majority of earlier studies that have been done on Rice plant diseases in Bangladesh but used smaller datasets and showed lower performance levels.