In this study, we created a semantic segmentation model to recognize the Cornu Ammonis 2 (CA2) area inside the horizontal hippocampal slice section in a microscopic transparent-light image, which was only recognizable by biomarkers such as immunohistochemistry. The U-Net was modified so that we could incorporate the way how experts recognized the CA2 area. We achieved 100% accuracy and 84% precision. We built a system on an edge computing device and provided a practical microscope solution to assist neuroscientists.