Optimized ResNet-50 Deep Learning Model for Highly Accurate Plant Disease Classification and Detection Efficiency

This research aims to develop a resilient and expert deep learning system to identify and classify plant diseases using ResNet (Residual Neural Networks). Such innovations will help overcome the complexities of disease identification under the challenging real field conditions and lead to improved accuracy of diagnoses in disease management for enhanced crop yields. A research architecture called ResNet-50 pre-trained using the ImageNet dataset was fine-tuned for the PlantVillage dataset-consisting of more than 54,000 healthy and diseased images in 38 classes. Data augmentation techniques such as rotation, flipping, and scaling were also applied for enhancing the robustness of the model. The training and validation of the model were performed in an 80-20 dataset split, and the performance evaluation was measured using accuracy, precision, recall, and F1. The ResNet-50 model achieved an accuracy of 98.7% on the test set. Thus, it outperformed conventional machine learning strategies and simple CNN architectures employed. High precision and recall values were achieved, especially for clearly distinguishable diseases with visual symptoms like tomato early blight and potato late blight. However, the precision and recall of the model were lower concerning diseases with similar or subtle symptoms; therefore, future works required further improvement. Results demonstrate that ResNet-based deep learning models effectively detect and classify plant disease very accurately. Here proposed systems provide very good practical endeavor automation towards disease diagnosis in agriculture, expecting huge reductions in crop losses. Future study will focus on improving the model's performance for less-symptomatic diseases and upscaling it for real-world field applications.