One of the most recent agricultural research topics is the recognition and classification of diseases from a plant leaf. With the exponential advancement of smart farming, plant disease detection becomes digitalized and data-driven, allowing advanced decision support, smart examination, and preparation. The detection of agricultural plant diseases using machine learning techniques would reduce the dependence on farmers to preserve agricultural goods. This paper proposes a deep learning-based metaheuristic algorithm of paddy leaf disease detection and recognition that enhances accuracy, generality, and training performance. This paper describes field images of various kinds of paddy leaf diseases: normal, bacterial blight, brown spot, and blast diseases. In this paper, the input image is assigned to pre-processing to remove artifacts and noise from the image. The Optimized Deep Convolutional Neural Network with Cuckoo Search (DCNN-CS) Algorithm is then used to classify leaf diseases by using the pre-processed image. During both the basic pre-training and fine-tuning phases of the DCNN approach, weights and biases are optimized using a cuckoo search algorithm (CS) to reduce classification errors. This DCNN-CS technique allows the application of simple statistical optimization methods with a reduced computing workload, resulting in high classification accuracy. Finally, the proposed DCNN-CS model's classification accuracy and efficiency were evaluated and compared to other Classification Techniques.