The outbreak of coronavirus illness (COVID-19), which has impacted approximately 195 Nations and dominions worldwide, is causing an unprecedented disaster for the world. Since the pandemic outbreak, computational model-based diagnostic methods have gained popularity as a means of assisting with COVID-19 case screening and diagnosis with the support of medical imaging proficiency considering CT scans, chest X-ray (CXR) scans, and MRI. Early research reveals that CT Images and CXR images from COVID-19- suffering victims exhibit anomalies of the correspond to specific radiological structures. Even for expertise radiologists, identifying these structures can be difficult as well as time taking. Using the transfer learning concept, we present a new convolutional neural network (CNN) supported by deep learning unification model and along with the long, short-term memory (LSTM) used. This framework extracts features from images by combining parametric quantity (weights) from multiple frameworks into a single framework, which then acts as a customized classifier for prediction. To display the affected regions of CXR and CT pictures, we employ gradient-weighted class activating mapping. In addition, the visual depiction of features helps better understand how separable the researched models’ classes are in terms of COVID-19 detection. The suggested model’s performance is measured by cross-validation studies on the datasets present in public that include both good health person and COVID-19 also other Bronchitis-infected CXR and CT pictures. According to evaluation results, the top-performing unification model has an advanced level of sensibility and particularity achieved a classification accuracy of 96.85%.