ResNet50 for Accurate Plant Leaf Disease Detection and Classification
Ritu Rani, Sheifali Gupta
Abstract
ResNet50 deep learning network in this research report obviously finds plant diseases. Divided into “Healthy,” “Powdery,” and “Rust,” the data compiles 1530 images overall and generates testing, training, and sets. Every image is modified and scaled using techniques of validation data augmentation aimed to increase variance. ResNet50 models gain from pre-trained weights to be multi-class classification focused. Changes in learning rate and early training phase stopping help to prevent overfits. To ensure good performance, the model is assessed with accuracy, precision, recall, F1-score, and confusion matrix. Plotting of training and validation accuracy and loss curves helps one to track learning development. After that, the model is ready for use with an easy interface; field testing is carried out to confirm its performance on fresh data. Consistent updates and active learning strategies help to guarantee ongoing development. This approach seeks to generate an accurate, scalable, efficient system for plant disease classification thereby improving agricultural output and sustainability.