The global population faced severe repercussions from the outbreak of the coronavirus (COVID-19) epidemic in late 2019, leading to a devastating health crisis and significant economic turmoil. As the pandemic persists, new strains continue to emerge, necessitating advancements in diagnostic techniques. Radiological image analysis using deep learning (DL) methods has proven effective in identifying infected regions. This study introduces a novel deep learning approach that combines convolutional neural networks (CNNs) and Darknet, uniquely tailored for accurate COVID-19 detection in chest X-rays and computed tomography (CT) images. Unlike previous methods, the proposed model integrates multi-stage categorization and segmentation techniques, enhancing its capability to precisely localize and identify COVID-19-infected regions. Results demonstrate that this innovative strategy achieves approximately 10% higher accuracy compared to recent state-of-the-art methods, with an accuracy of 98.06% in binary classification. By leveraging this original methodology, the model holds significant potential to revolutionize rapid and reliable COVID-19 diagnostics through radiological imaging.