IMCCRT 2025 - 4528 Title - Modified Densenet-201 Architecture: A Deep Learning - Driven Rice Blast

Title - Modified Densenet-201 Architecture: A Deep Learning - Driven Rice Blast Detection

Author's Name - Marionne J.F. Tapado, Melojean C. Marave

Abstract -Rice is a crucial food source for more than half the world’s population and plays a key role in Asia's economy. However, diseases can significantly reduce crop yields, leading to financial losses and food shortages. In Zambales, rice diseases could decrease farmer incomes and reduce the local rice supply, impacting food availability in the region.
This study focuses on improving a DenseNet-201 neural network to accurately detect rice blast disease, enabling farmers to better manage crops and reduce losses. By offering a reliable tool for detecting rice blast disease, this research contributes to sustainable rice farming practices.
The modified DenseNet-201 architecture in this study incorporates global average pooling and reduces the size of the fully connected layers in the original DenseNet-201 model. The data used in this study consists of images of healthy and rice blast-affected leaves. The training and validation datasets were sourced from Kaggle, while the testing data was collected from Masinloc, Zambales, Philippines. These images were annotated and validated by three agricultural experts.
The Modified DenseNet-201 model showed outstanding performance, achieving impressive results across multiple evaluation metrics. The model reached an accuracy of 99.50%, precision of 100%, and both recall and F1-Score values of 99%. These results indicate that the model is highly effective in accurately classifying rice leaf diseases.

Keywords: Algorithm Enhancement, Deep-Learning, DenseNet-201, Rice Blast Disease.