Research in agriculture, specifically on paddy plants, is needed to accurately categorize the diseases that affect rice crops in their early stages. This is achievable if automated technologies are capable to support farmers in identifying rice diseases based on photographs of paddy leaves. Applying deep learning (DL) techniques to identify diseases in farmed plants can reduce farmers dependency on the protection of paddy crop production. In this study, we compiled a dataset comprising five distinct classes: bacterial leaf blight, bacterial leaf streak, bacterial panicle blight, downy mildew, and normal. We leverage a variety of DL models to check the system validity in the plant disease detection field. Our hybrid model, CNN-ResNet50-BiLSTM, exhibited exceptional performance, boasting an extraordinary training accuracy of 98.67% and a validation accuracy of 99.93%. On top, we evaluate the precision, recall, F1 score, and anticipated results to determine the effectiveness of the proposed system. This promising system has the potential to support agricultural professionals in the early and precise diagnosis of paddy leaf disease, ultimately improving crop outcomes and alleviating the burden on agricultural systems.