The deadly disease of pneumonia is a global problem, affecting many people. As a result, it threatens lives and burdens healthcare. Recently, deep learning techniques have gained traction in healthcare applications to assist in automating disease detection. This paper focuses on creating a system for the automated identification of pneumonia from chest X-ray images. We use a modified UNet model with optimized hyperparameters to achieve superior results. We compare DenseNet and UNet to evaluate performance. Our model achieved 84.78% accuracy on the test dataset. The model also demonstrated a strong recall rate of 86.78% and an F1 score of 87.99%. Our findings highlight machine learning’s role in speeding pneumonia detection. Our model explores the utility of UNet in such a domain, offering a new approach that builds on fundamental methods in medical imaging. The findings indicate that basic models can be successfully utilized in clinical diagnostics, a field that has been relatively underexplored until now.