How nada-ben-ali/covid19-detection-using-federated-learning works
This repository combines a federated-learning COVID image classifier with a web interface for clients and training operators. The client workflow is: submit an image for prediction or respond to a training invitation; accepting an invitation starts client training. The operator workflow launches a federated session, whose server aggregates participant updates and evaluates the global model. The graph includes the sampled client API, model preparation and server strategy, plus major frontend and preprocessing modules indicated by the source index. Some backend and UI implementation details were not sampled, so uncertain links are omitted.
Main entry points: App.js Route application · main.py Client API · server.py Session orchestrator · FLstrategies.py Aggregation strategy
Components 20
Web interface
- Route application
front_v2/src/App.js - Sign in
front_v2/src/pages/Signin.js - Sign up
front_v2/src/pages/Signup.js - Navigation
front_v2/src/components/Navbar.js - Client workspace
front_v2/src/pages/client/Client.js - Server workspace
front_v2/src/pages/server/Server.js - Notification widget
front_v2/src/components/notif.js
Client services
- Client API
client/main.py - Training participant
client/client.py
Federated training
- Server API
server/server_api.py - Session orchestrator
server/server.py - Aggregation strategy
server/FLstrategies.py
Data and models
- Client model utilities
client/preprocess_model.py - Server model utilities
server/preprocess_model.py - Session weights
fl_sessions - Training configuration
config_training.json - Client data partitioning
Preprocess/split_data_clients.py
Other components
- User
- MySQL database
- Flower runtime
Connections 18
- User → Route application: opens interface
front_v2/src/App.js - Route application → Sign in: routes to
front_v2/src/App.js - Route application → Sign up: routes to
front_v2/src/App.js - Route application → Client workspace: routes to
front_v2/src/App.js - Route application → Server workspace: routes to
front_v2/src/App.js - Route application → Navigation: renders
front_v2/src/App.js - Client workspace → Client API: sends requests
front_v2/src/pages/client/Client.js - Client API → Client model utilities: loads model
client/main.py - Client API → MySQL database: queries and updates
client/main.py - Client API → Training participant: starts on acceptance
client/main.py - Training participant → Client model utilities: imports utilities
client/client.py - Server workspace → Server API: requests session data
front_v2/src/pages/server/Server.js - Server API → Session orchestrator: imports server
server/server_api.py - Session orchestrator → Aggregation strategy: creates strategy
server/server.py - Session orchestrator → Flower runtime: starts server
server/server.py - Aggregation strategy → Server model utilities: builds model
server/FLstrategies.py - Aggregation strategy → Session weights: saves aggregated weights
server/FLstrategies.py - Session orchestrator → Training configuration: reads session settings
server/server.py