Smart agriculture guarantees food production’s sustainability through innovative technologies to monitor and control plant health. Early and effective identification of plant diseases is fundamental for increasing the quantity and quality of agricultural products. This paper introduces a new convolutional neural network (CNN) architecture to identify plant diseases. We trained the proposed model using a comprehensive dataset called the PlantVillage dataset. The suggested CNN achieved an accuracy of 99.55%. This accuracy greatly surpasses the accuracies of pre-trained models like ResNet50, EfficientNetB2, and Xception. Comprehensive assessments demonstrate that the proposed model is more efficient regarding the number of parameters and storage space, processing time, and computation cost. This research enhances the field of agricultural technology by offering a robust tool for the prompt detection of plant diseases, which has the potential to increase crop productivity and minimize losses.