In various applications, deep neural networks has produced impressive accuracy, though the success rate is often attributed on heavy network architectures. In this way, a portable student network with significantly fewer parameters can achieve considerable accuracy, compared to a teacher network. But, in addition to the fact that the accuracy, reliability, and timeliness of the learning student network is also important for practical use. In this paper, we make the model to predict the maximum efficiency using the combination of student-teacher network to predict the occurrence of COVID-19 in a person. Three different architectures are used and compared to find out which one gives the highest accuracy. A web application is developed as a frontend for displaying the test results. Here, the different combination of blocks are trained in parallel and finally concatenation of the blocks are done to improvise the efficiency of the model which in turn predicts the presence of COVID-19 among the patients effectively with an accuracy of about 90% at low cost.