Umme Sadia Raisa2025-11-232025-11-232024-05-13https://dspace.nstu.ac.bd/handle/123456789/153In the rural landscape of Bangladesh, where rice reigns supreme as the cornerstone of sustenance, disease poses a formidable threat to the nation's agricultural backbone. With a staggering 78% of the country's arable land dedicated to rice cultivation, the specter of crop diseases looms large, exacting a heavy toll on farmers' livelihoods and the nation's food security. Efficient disease detection is paramount to safeguarding rice yields and ensuring food security. Traditionally, this task has been labor-intensive and time-consuming, demanding extensive manual effort. However, the advent of automated systems offers a beacon of hope in this agricultural conundrum. Hence deep learning is getting popular for its transformative technology rapidly eclipsing conventional approaches with its unparalleled performance. The objectives of this research were twofold: to achieve higher recognition accuracy for rice diseases through transfer learning and to compare the performance of the three CNN architectures. The study found that VGGNet16 and InceptionResNetV2 models, when optimized with RMSprop, achieved high accuracy rates of 98.44% and 98.21%, respectively, demonstrating their robustness and reliability for rice disease detection. In contrast, ResNet50 showed lower performance with an accuracy of 50.99%, indicating its limitations in this specific application. This research contributes by demonstrating the effectiveness of transfer learning in improving disease detection accuracy and providing a comparative analysis of CNN models and optimizers. The findings highlight the potential of these models for broader agricultural applications, offering insights into crop health management beyond rice cultivation. The study underscores the importance of selecting appropriate models and optimization strategies to enhance the efficiency and accuracy of automated disease detection systems, thereby supporting agricultural sustainability and food security. KEYWORDS: Rice; Disease Classification; CNN; Deep Learning; Transfer Learning; VGGNet; ResNet50; InceptionResNetV2; Feature Extraction;