The risk of lung disease is immense for many people, espe- cially in developing countries, where billions of people face energy poverty and are dependent on polluting forms of energy. The World Health Organization estimates that more than four million premature deaths occur each year from diseases related to household air pollution, including pneumonia. Radiologists diagnose and detect medical conditions with imaging techniques such as CT, MRI, and X-rays. However, they face many challenges interpreting chest radiographs in high workload conditions, even for highly experienced physicians. A tool for automatically locating and classifying anomalies would be of great value, and a deep learning approach provides several ways to achieve this goal. In this study, we train Faster R-CNN neural network for lung disease classification using the feature extraction networks such as ResNet, CheXnet, and Inception ResNet V2. The experiments are conducted on a dataset of 112,000 images with corresponding labels annotated by experienced radiologists. The experimental results show that the models can identify the exact lesion area for a given chest X-ray and the classification accuracy is up to 95.5%. The Grad-CAM is performed to highlight the lesion area thus reducing stress for physicians while providing patients with a more accurate diagnosis.