How k-gokulappadurai/respirenet-respiratory-disease-prediction-web-application-using-deep-learning works
This repository predicts respiratory conditions from sound recordings and also trains the classifier. The primary user workflow is to enter a WAV path in the Kivy GUI, extract padded MFCC features, run the saved Keras model, and display class scores and the top prediction. A separate training workflow prepares labeled recordings, trains a CNN, saves a model, and evaluates it. The README also describes batch prediction. The output files appear to be captured console sessions rather than runtime modules; their roles are inferred from their excerpts. No source shows a Flask application in this repository.
Main entry points: main.py GUI inference · respiratory-disease.py CNN training · resp_model_300.h5 Saved classifier · result of model.py Batch prediction
Components 12
Audio input
- Kivy GUI
main.py - Audio features
main.py
Prediction
- GUI inference
main.py - Batch prediction
result of model.py - Saved classifier
resp_model_300.h5
Training and evaluation
- Labeled recordings
- Training features
respiratory-disease.py - CNN training
respiratory-disease.py - Model evaluation
respiratory-disease.py
Run artifacts
- Training transcript
output.py - Prediction transcript
output2.py
Other components
- Audio user
Connections 10
- Audio user → Kivy GUI: enters audio path
README.md - Kivy GUI → GUI inference: invokes prediction
main.py - GUI inference → Audio features: extracts MFCCs
main.py - GUI inference → Saved classifier: loads model
main.py - GUI inference → Kivy GUI: returns scores
main.py - Batch prediction → Saved classifier: loads model
result of model.py - Labeled recordings → Training features: provides recordings
respiratory-disease.py - Training features → CNN training: supplies MFCCs
respiratory-disease.py - CNN training → Saved classifier: saves model
respiratory-disease.py - CNN training → Model evaluation: evaluates predictions
respiratory-disease.py