The widespread COVID-19 outbreak had a negative influence on many people’s health. The COVID-19 virus infected 305.9 million individuals on January 10, 2022. COVID-19 RT-PCR testing may be replaced by computed tomography (CT) images. The Unet model can detect ROI in coronavirus-generated CT images. During early COVID-19 segmentation, the Going Glass Opacity (GGO) pipeline resembles a healthy lung. COVID-19’s suggested “convUnet” model can efficiently segment the GGO. The investigation made use of the COVID-19 CT scan lesion segmentation dataset. We outperformed the Unet and other current models on numerous metrics. Infection with COVID-19 is difficult to detect. COVID-19 is presently being studied with the use of medical imaging and computer vision. Available datasets have limited medical images and internal variability due to varying medical imaging technology, methods, and experiences. Using the suggested approach, it detects COVID-19 in lung CT scans. Many datasets are received and evaluated, and domain adaptation is employed to create domain-invariant representations of medical images. To create comparable representations of medical images, the triplet loss function must be applied (infected cases). The suggested model diagnostic approach is compared to existing COVID-19 diagnostic methods. This method enables clinicians to diagnose COVID-19 with greater accuracy, precision, and dependability. These COVID with excellent the proposed method performance was tested against other approaches. The proposed model achieves significant classification increases, up to 92.6%, even when the dataset is not trained on the same one.