Anomaly Detection project in ECG signals - Part 2

I have built a web application to detect anomalies from ECG signals. It takes as input a CSV file that will be used to evaluate the model. After the 'Check Anomalies' button is pressed, the app will display a plot to visualize the true values versus the predictions obtained the deployed model. The green points represent the predicted label, while the red crosses represent the ground truth.

Detailed description of the project

The article with the explanations is An End to End Anomaly Detection App for ECG signals with DagsHub, SageMaker, and Streamlit. You can also find link to my deployed app is here.

Tools used in the project

• DVC
• DagsHub
• MLflow
• AWS Lambda
• Streamlit

Project Structure

• ecg_data/: contains all the data, training and test data
• if/: contains the artifact of the isolation forest
• doc/: documentation
  • set_environment.md: Instructions to set up the environment in your local PC
  • create_datapipeline.md: Command lines to create the data pipeline
• src: contains the following scripts
  • train.py: Python script to track the experiments of the ML model and save the model artifact on MLflow's platform
  • deploy.py: Python script to deploy MLflow model with AWS SageMaker
  • ecg_app.py: Python script to create the web application with Streamlit

Split data into training and test set

python src/create_data.py

Train and evaluate model

python src/train.py

Deploy MLflow Model to a Sagemaker Endpoint

mlflow sagemaker build-and-push-container

Deploy model to SageMaker

python src/deploy.py