PLANT DISESASE DETECTION

Rice diseases caused by bacteria, viruses, and fungi significantly impact global rice production. Early and accurate identification is crucial for maintaining food security. Image analysis offers a promising solution, but existing AI models struggle with unseen data. This research addresses this challenge by proposing a novel CNN model for rice disease recognition with a reduced number of parameters, making it more efficient. The model was trained on a unique dataset of 4,199 rice leaf disease images encompassing five common diseases. It achieved outstanding performance, boasting a training accuracy of 99.78% and a validation accuracy of 97.35%. Even more impressive, the model maintained high accuracy (97.82%) on a completely independent dataset, demonstrating excellent generalizability. Furthermore, the model outperformed existing CNN models in binary classification tasks, achieving recognition rates of 97% for Blast, 96% for Brownspot, and 96% for Bacterial Leaf Blight. These results solidify the effectiveness and superiority of this new approach for rice diseaserecognition compared to current state-of-the-art methods.