Utilizing Transfer Learning for Crop Suggestions and Plant Disease Identification

Environment changes have brought about a rising measure of unforeseen precipitation, standard beneath temperatures, and heatwaves in the district, bringing about a critical loss of the biological system. Machine Learning has created different utility devices to handle world issues. This issue of farming can be addressed by utilizing various ML algorithms. This study addresses two objectives: a) recommending crops based on diverse environmental and soil parameters, and b) utilizing machine and transfer learning techniques to detect plant diseases. The datasets utilized in this research were publicly available online. The dataset underwent training using eight distinct algorithms: Decision Tree, Random Forest, Multi-Layer Perceptron, XGBoost, Logistic Regression, Naive Bayes, Support Vector Machine (SVM), and K-Nearest Neighbors. For the second task, eight CNN architectures—VGG-19, ResNet50, Efficient-NetB0, EfficientNetB6, DenseNet169, Xception, InceptionV3, and MobileNetV2—were trained, followed by a comparative analysis between the two tasks. Random Forest achieved an accuracy of 99.31% for the first task, while ResNet50 attained 99.2% accuracy for the second task.