A ConvNet trained on a Kaggle Dataset that takes chest X-ray as an input and predicts whether the patient has pneumonia or not. Integrated it into a web app with Flask and deployed it on Heroku.
Achieved 0.83 recall, 0.91 precision and 0.87 f1-score. The model used to overfit a lot, so used data augmentation, batchnorm and dropouts to reduce variance. Used this GitHub repo as a reference and modified it to suit the model and the website.
Link to the website: https://pnclassify.herokuapp.com/
- DL Model : Python, TensorFlow, Cv2
- Backend : Flask
- Frontent : HTML, CSS, Bootsrap, JavaScript
- Deployment : Heroku