The viral outbreak of COVID-19 that started in the year 2019, radically changed our everyday life, with a detrimental impact on the simple, daily habits of citizens. In many countries around the world, the usage of mask is necessary as a protection measure against covid-19. Every service, organization, various stores, schools, universities, hospitals, companies and many other places, which are attended by hundreds of people every day, make the use of a mask necessary to enter them. This fact requires the control of the persons when they enter the respective spaces to determine if they are wearing a mask when entering the area. In this research we compared performance on YOLOv4 and the Tiny-YOLOv4 algorithm on images, recorded video, and real time video. In the next step we will implement the YOLOv4 TFlite and Tiny YOLOv4 TFlite model for mobile applications using the Android Studio platform. On the proposed dataset YOLOv4 achieved 92.91% mAP and training took around 2 hours for 1000 iterations. On the other hand, YOLOv4-tiny achieved 74.75% mAP and training took less than half an hour for 1000 iterations. For further improvement we convert YOLOv4 and YOLOv4-tiny to YOLOv4 TFlite and YOLOv4-tiny TFlite respectively. After this step we compare YOLOv4 TFlite and YOLOv4-tiny TFlite model performance on mobile device. YOLOv4 TFlite achieved 96.92% accuracy on real time video at 5017ms and YOLOv4-tiny 74.72% accuracy on real time video at 491ms.