With the worldwide spread of the COVID 19 virus, the statistics show that over 5.95M of people in the world are positive cases, while 365K of them were dead, over these last four months, and this is due to the speed of infection resulting from this epidemic, which occurs through the respiratory tract of humans. The government makes some rules to maintain human health, e.g., wearing respiratory protective devices, respecting safety distance between people, and so on. Unluckily, many people don’t respect the rules, and it’s so hard to know if they are respecting the rules or not. To tackle this problem, and respect the social distancing in the same, robots were used for checking temperature, detecting face mask, and so on. In this thesis, I worked on detecting face mask problem, and to do this I used deep learning, which is the state-of-the-art method, to perform object detection and image classification, generally, in deep learning, big dataset is required to get better results, as coronavirus is a new virus, we don’t have a large dataset for masked faces, so in our work, we study the use of machine learning with small training dataset, we make experiments with different Deep Learning architectures over a small balanced and unbalanced dataset, we propose new models and compare them with the state-of-the-art models, and according to our experiments, we prove the high performance of our models on the used dataset.