Glasses were originally only a visual aid for someone who had a visual impairment, but over time glasses have developed into a fashion necessity. Glasses can generally be tried directly in optics, but due to the pandemic, there are restrictions on interaction so that glasses cannot be tried directly. This research discusses face shape recognition using the Viola Jones method and the Convolutional Neural Network method which is useful for providing recommendations for selecting glasses models via online. The input data is an external data from the Kaggle site which has five face shapes namely heart, rectangle, oval, round, and square. The training process is carried out to train the machine to recognize the user's face shape according to its class. The testing process provides accuracy results of 84.38% and macro average values for precision of 85%, recall of 85% and F1-Score of 84%. This system is expected to help users of glasses to choose a model of glasses that suits their face shape online.