Recognition is one of the many problems encountered today, this problem has several ways to be solved. This research used Convolutional Neural Networks (CNN), which is a deep neural networks method as a means of face recognition, which has been proven to be widely used in face classification, using a dataset of male and female facial photos totaling 27,167 photos, of which 17,678 are male and 9,489 are male. woman. To avoid unbalanced data processing, the researchers disguised the photos of women and men so that the total photos used for the training amounted to 18,978 photos. Besides that, the researcher also added dropout as a test parameter. The author uses python to implement gender differences in the images in the data that has been prepared. For the preparation of the Convolutional Neural Networks model architecture the authors use several layers. Then the data will be trained before being tested with new data that has been prepared where the new data for testing is divided into two datasets to see if there are differences in accuracy results. What distinguishes the two datasets is the position of the photo and the background of the photo. Of the two existing datasets, the first dataset produces an average of 73.33%, while the second dataset produces the highest 84.34%.