A Deep Learning Framework for Early Detection and Diagnosis of Plant Diseases

The paper explores the application of deep learning techniques to the early detection and diagnosis of plant diseases, aiming to enhance accuracy and reduce the response time compared to traditional methods. Utilizing a hybrid model that combines the strengths of ResNet-50 and Inception-v3 architectures, the study leverages a comprehensive dataset of augmented plant images, categorized into several disease states and healthy conditions. The dataset was processed using a series of image augmentation techniques to mimic various real-world conditions, improving the model’s robustness and generalizability. The best-performing model is the hybridized model of ResNet-50 and Inception-v3 which has achieved 0.97 in accuracy and recall, and 0.96 in F1-score on the validation set, showcasing its potential to significantly aid in early disease detection efforts. Challenges such as class imbalance and overfitting were addressed through techniques like class weighting and dropout layers, respectively. The findings suggest that the hybridized deep learning architecture can be a powerful tool in the agricultural sector, providing rapid and accurate diagnostics capabilities that could be particularly beneficial in preventing widespread crop damage. The work aims to focus on scaling the solution to handle more diverse plant species.