1. AML Dashboard

The AML Dashboard serves as a user-friendly tool for training, evaluating, and applying AML models to several datasets. Built upon the Marcelle framework, the dashboard offers standard interaction GUIs and the ability to create custom interactive widgets compatible with Marcelle’s structure.

The primary goal of the AML Dashboard is to enable users to:

  1. Collect personal gesture data or load existing datasets.

  2. Train classifiers using the AML Engine.

  3. Compare the performance of AML models with other ML approaches.

To achieve this, the AML Dashboard integrates seven main interfaces:

  • Dataset Interface: Allows users to collect data through personal gestures or load existing datasets.

  • Training Interface: Facilitates the training of AML models based on loaded datasets.

  • Model Exploration Interface: Enables users to explore and compare the capabilities of AML models with other ML approaches.

  • Real Time Exploration Interface: Applies the trained AML algorithm in real-time. This feature is only available when the dataset chosen is Sensors. It enables users to explore new sound patterns using personal gestures.

  • Model Fetching Interface: Fetches trained models and statistics using an AML-IP Collaborative Learning scenario.

  • Context Broker Interface: Interacts with the Context Broker to handle inference data.

  • AML-IP Management Interface: Manages the AML-IP nodes, allowing users to start, stop, and delete the nodes as well as monitor their status.