From X-Ray to Diagnosis: How Transfer Learning Powers Pneumonia Detection at Scale

With over 2 million pneumonia deaths annually, particularly in low-resource areas with limited access to radiologists, automated diagnostic tools are critically needed. This project demonstrates how transfer learning with MobileNetV2 can achieve 90%+ accuracy on real clinical data, making medical AI accessible through a free, deployable model—though challenges like false positives on non-medical images reveal the complexity of production medical AI.