Skip to content
H
All projects

AI / ML · ML Engineer

Pneumonia Predictor

A full-stack deep learning system for pneumonia detection from chest X-ray images, built with PyTorch and Keras/TensorFlow and served via a Flask REST API with a responsive single-page web UI.

View repository

Highlights

  1. 01

    Built a 4-model deep-learning ensemble for pneumonia detection (ResNet18, ResNet50, VGG16, EfficientNetB0) across PyTorch and Keras/TensorFlow, combining per-model probability outputs via majority voting with a conservative NORMAL-bias tie-break appropriate for clinical screening.

  2. 02

    Trained on 5,200+ labeled chest X-rays (Kaggle dataset) via transfer learning and fine-tuning across both frameworks, exposing evaluation through /predict, /evaluate, /confusion-matrix, and /health Flask REST endpoints.

  3. 03

    Shipped an end-to-end inference service — Flask REST API + a zero-dependency vanilla HTML/CSS/JS single-page UI with drag-and-drop upload, live preview, animated confidence bars, and per-model vote breakdown — runnable in one command (python start.py).