Early and accurate detection of COVID-19 using automated techniques is a key challenge in public health. This paper presents a Deep Learning model based on the Xception architecture, trained with a set of chest X-rays from patients positive and negative for COVID-19. The model was optimized using regularization techniques (Batch Normalization and Dropout) and evaluated using data partitioning strategies for training, validation, and testing. The results show outstanding performance, achieving high accuracy and sensitivity compared to previous literature studies, validating the effectiveness of the proposed approach. Likewise, a web-based platform was implemented that allows the uploading and analysis of diagnostic images in real time, tested with 15 real radiographs, confirming the applicability of the system in clinical settings supporting diagnosis. This contribution combines technological innovation with a practical approach, providing a scalable and easily accessible tool for the automated detection of COVID-19 from medical images.