Crop diseases significantly threaten global food security, impacting agricultural productivity and economic stability. These diseases are caused by various pathogens, including fungi, bacteria, viruses, and nematodes, which can infect various plant parts, including leaves, stems, roots, and fruits. Classifying the several crop diseases is the requirement for the prevention of distinct disease problems. However, it is challenging to detect exact crop diseases that cause slight differences among the diseases of the same crop. Meanwhile, multi-layer convolutional neural networks, while effective in daily computer vision tasks, come with drawbacks such as significant computational memory requirements and extended training times. The simplified convolutional neural network (SCNN) model comprises three hidden layers with increasing order in each convolution kernels 16, 16, and 32, reducing the time and space complexity. This study incorporates normalization, dropout, and regulation techniques to accelerate training merging and enhance accuracy. Then, the performance metrics are found, and distinct algorithms are compared to measure the effectiveness of the top-performing model. The investigational comparisons among the projected SCNN model and others revealed that the planned SCNN model offers the uppermost accuracy. Furthermore, the SCNN outcome is applied to actual crop image datasets, achieving a classification accuracy of 95.69%. Above all, the planned SCNN model demonstrates promising results in potato blight disease classification, offering high accuracy while mitigating the computational memory requirements and training time. These findings suggest its potential applicability in real-world agricultural scenarios for efficient crop disease detection and prevention.