The objective of this thesis was to design and implement a face recognition application based on a convolutional neural network. The neural network used by the application was required to distinguish persons apart based on the images of their faces. A readily available neural network architecture was chosen, and transfer learning was used in the training process. To achieve the learning goal, a new loss function was introduced by the author. The use of the new loss function enabled a sufficient accuracy of the demo application. The code of the application was written in Python. The learning process of the model was stable, and a moderate accuracy was reached. Signs of overfitting during training were not detected. The resulting application was simple, but it provided relatively accurate predictions considering the short training time, small size of the model and the constricted source data.