Atomic Force Microscopy(AFM) can reveal the upper surface of biomolecules such as proteins at a resolution of less than 1nm. It doesn’t describe much about the individual atoms, but it gives a good approximation of the surface. So, combining this high resolution surface information with low resolution 3-Dimensional information obtained through x-ray crystallography can give a more accurate and complete 3-Dimensional information of the biomolecule structure. However, the image obtained through AFM has only some surface information, not the whole, and the shape of biomolecules is very similar. Therefore, unless the user is adept technician it is difficult to know exactly what biomolecule it is and which part of biomolecule it is. In this study, we have studied to classify amino acid AFM data automatically by using deep learning which shows high performance in image classification and segmentation. The training data was simulated by the AFM scan method of the program Microscope Simulator 1.3.1, and about 400,000 random amino acid AFM scanning images were generated through simulation. The network using VGGNet was used for learning, and as a result, accurate amino acid AFM image classification was possible with accuracy of about 93.5%.