- In this project two models ANN and CNN were built from scratch using TensorFlow and Keras
- The problem statement is to detect whether patient is having pneumonia or not by training models with x-ray images
- Data Augmentation was performed and the main reason behind it is to provide maximum randomness to the training data where we use a function to shuffle images, make minor changes, and save images accordingly because it is a good practice to provide as much randomness as possible in order to achieve good accuracy on real-time data and testing data
- Separable Convolutional layers were used which proved to be more robust and powerful
- After Comparing the accuracies of both the deep learning models best model was selected for further deployment on android application
- The model was converted in .tflite format in order to deploy it using TensorFlowlite dependencies in flutter
- UI of app was done using Flutter and Dart while functionalities were implemented using TensorFlowlite libraries
- The app enables user to click an image of x-ray or select the image from app storage or drive to check the results
- UI is configured in such a way that the result is displayed on the screen
- If patient has pneumonia a short and useful prescription from experienced doctors is made available in the app itself
- The models were built on Google Colab by fetching data through Kaggle API
- tflite: ^1.0.2
- image_picker: ^0.8.4+4
- flutter_svg: ^0.23.0+1
