COVID-19 has caused a global health crisis that has infected millions of people across the globe. Currently, the fourth wave of COVID-19 is about to be declared as Omicron. The new variant of COVID-19 has caused an unprecedented increase in cases. According to World Health Organization, safety measures must be adopted in public places to prevent the spread of the virus. One effective safety measure is to wear face masks in crowded places. To create a safe environment, government agencies adopt strict rules to ensure adherence to safety measures. However, it is difficult to manually analyze the crowded scenes and identify people violating the safety measures. This paper proposed an automated approach based on a deep learning framework that automatically analyses the complex scenes and identifies people with face masks or without facemasks. The proposed framework consists of two sequential parts. In the first part, we generate scale aware proposal to cover scale variations, and in the second part, the framework classifies each proposal. We evaluate the performance of the proposed framework on a challenging benchmark data set. We demonstrate that the proposed framework achieves high performance and outperforms other reference methods by a considerable margin from experimental results.