So, what is it about?
The current workflow handles model training and evaluation in isolation, with no integrated real-time inference or user interface.
Implement an end-to-end pipeline where incoming ECG signals are pre-processed, passed through the trained model, and used for live arrhythmia detection and alerting.
Technical Notes:
Build a unified inference pipeline connecting signal acquisition → pre-processing → model inference → output handling.
Load the best saved model automatically from checkpoints for real-time predictions.
Integrate the existing pre-processing module to ensure consistent data formatting.
Implement logic to track 3–4 consecutive abnormal predictions and trigger a notification or warning event.
Use a lightweight interface framework such as Streamlit to visualize input signals, predicted classes, and alerts in real time.
Expected Outcome:
Fully automated signal-to-prediction workflow.
Dynamic loading of the best available model.
Real-time notification on persistent arrhythmia patterns.
User-friendly monitoring dashboard for live signal and prediction visualization.
Code of Conduct
So, what is it about?
The current workflow handles model training and evaluation in isolation, with no integrated real-time inference or user interface.
Implement an end-to-end pipeline where incoming ECG signals are pre-processed, passed through the trained model, and used for live arrhythmia detection and alerting.
Technical Notes:
Expected Outcome:
Code of Conduct