A key component of world agriculture, paddy farming serves as the main source of food for a sizeable portion of the world’s population. However, a number of diseases may strike paddy crops, which might have a detrimental impact on output and food security. Traditional disease identification techniques are often described as time-consuming and labor-intensive. In this context, deep learning—more specifically, Convolutional Neural Networks (CNNs)—has shown promise for expediting the detection process. An improved CNN-centric approach to the automatic detection of paddy illnesses is presented in this study. The suggested approach leverages architectural improvements, hyperparameter optimization, transfer learning, and dataset augmentation to integrate the latest developments in CNN architecture. The goal of these modifications is to increase the model’s capacity to recognize visual cues linked to various paddy illnesses. The model may be trained on a variety of enhanced datasets to make it more resilient and able to generalize in a range of environmental circumstances. Even with a small amount of labeled data, the CNN may be more easily adapted to the particular job of paddy disease detection using transfer learning from pre-trained models on big picture datasets.