Rice is a staple crop worldwide, providing sustenance to over half of the global population. Like other crops, it is vulnerable to a range of diseases that can severely reduce both yield and quality, sometimes resulting in losses ranging from 20 to 40%. Early disease detection is critical for preserving high yields, requiring farmers to recognize various diseases and their visual symptoms. However, regularly monitoring large agricultural areas is often impractical and costly, which would ultimately raise rice prices for consumers. A potential solution to this challenge involves utilizing machine learning algorithms integrated with drone technology and the Internet of Things to monitor crop health efficiently and detect diseases promptly. In this paper, we introduce a deep convolutional neural network that employs optimization techniques such as stochastic gradient descent with momentum, adaptive moment estimation, and root mean squared propagation, as well as learning strategies like transfer learning and baseline learning, to accurately detect and classify rice leaf diseases. This study also explores the use of ensemble learning, leveraging ResNet-50 and ResNet-34 architectures for rice disease detection. We apply transfer learning and baseline learning, including fine-tuning, to tailor these ensemble models to the specific task of disease identification. The models are initially trained on a rice disease dataset, with the output layers modified to classify 11 different disease categories. The effectiveness of this ensemble approach is evaluated, demonstrating its potential for accurate and efficient disease identification in rice crops.