Pneumonia Detection from Chest X-ray using CNN

The Pneumonia Detection system is a healthcare-AI final-year project that classifies chest X-ray images as normal or pneumonia-positive using a fine-tuned convolutional neural network. The model is trained on the NIH Chest X-ray dataset (112,000+ frontal-view images) and fine-tunes a ResNet50/MobileNetV2 backbone pre-trained on ImageNet — the transfer-learning story alone answers a dozen viva questions. The LungScan Streamlit app returns the prediction with a confidence score and overlays Grad-CAM heatmaps on the lung regions that drove the decision, so the examiner sees visual proof, not just a number. A batch-screening dashboard lets you process a folder of X-rays and export results for your report's evaluation chapter.

This is a built-to-order software project: code, trained model weights and the complete viva kit (report, PPT, Q&A) are prepared fresh for you.

1. A chest X-ray is uploaded to the LungScan app (or dropped into the batch folder).
2. The preprocessing pipeline resizes and normalizes the image and applies CLAHE contrast enhancement so lung opacities stand out.
3. The fine-tuned CNN (ResNet50/MobileNetV2) extracts deep features and outputs a pneumonia probability.
4. Grad-CAM back-projects the prediction onto the X-ray, highlighting the infiltrated lung regions in a heatmap.
5. The dashboard shows the verdict, confidence and heatmap per image, and aggregates batch results into the metrics tables used in your report.

• Normal vs. pneumonia binary classification of chest X-rays with confidence score
• Transfer learning on ResNet50/MobileNetV2 pre-trained backbones
• Grad-CAM heatmaps overlaid on lung regions as visual evidence
• LungScan Streamlit app — upload, predict and visualize in one click
• Batch-screening dashboard: process folders of X-rays, export CSV results
• Class-imbalance handling (weighted loss / augmentation) documented in the report
• Training notebooks with ROC curves, confusion matrix and per-class metrics
• CLAHE contrast enhancement and lung-aware preprocessing pipeline
• Evaluation chapter: accuracy, sensitivity, specificity, F1 and AUC

• Complete source code (training, evaluation, inference, Streamlit app)
• Trained model weights (.h5) — demos run on any laptop CPU
• Data preparation scripts and augmentation pipeline
• Project report PDF (clinical background, literature survey, methodology, results, ethics)
• PPT presentation for final review
• Viva Q&A preparation document (CNNs, transfer learning, Grad-CAM, ROC/AUC, class imbalance)
• Setup guide (environment, dependencies, dataset download steps)