Introduction: Covid-19 is a novel virus that has exponentially increased the number of infected persons and the death of human beings into millions within a few months. This virus spreads when a person comes into contact with another person, coughing, sneezing and droplets. Objective: To avoid loss of lives, human direct assistance and early precaution, automated systems required for reducing the number of cases. Deep Learning can facilitate human life much better way by automating human visual intelligence into machine intelligence. Methods: In this novel research work, we are implementing transfer learning methodology to improve the learning of a related objective task on top of base deep learning model in developing a mask/non-mask detection model along with changing the hyperparameters and data augmentation technique by using less input dataset for smart healthcare, smart home in reducing and detecting corona cases. Results: We used object detection model Single Shot Multi-box Detector and classification model mobile net, which achieved significant accuracy and much faster for both training and inference with prediction accuracy of 87% with IOU=.75 on our own created trained dataset comparable with other real-time object detection model such as Faster Regional Convolutional Neural Network by tuning the hyperparameters. Conclusion: The automated system not only reduces the false alarm but also enhanced the performance accuracy by detecting the mask and non-mask due to which the number of covid-19 cases can be reduced at an early stage.