Skip to main content

Overview

Introduction​

The Models page allows users to create, configure, and manage machine learning models within Clarity. Users can view model versions, training configurations, and performance metrics, as well as manage test predictions and deployments.

Clicking a model opens the Model Details Panel, which includes the following tabs:

  • Overview – Displays model metadata, settings, and training controls
  • Test Predictions – Used to evaluate model performance on unseen data
  • Compare – Enables side-by-side comparison of different model versions
  • Deployments – Used to deploy trained models for production inference

Model Details Panel:​

The Overview tab provides key model configuration tools and summarizes essential metadata and version details.

Status, Version, and Metadata​

Displays the current model version, status, ID, last modification date, and author.

  • Use the outlined version button to switch versions or start a new one

Settings Panel​

The Settings section defines the key parameters used to configure a model prior to training. These must be correctly set before initiating training.

Model Type:
Select the type of model to train: Classification, Unmixing, Target Detection, or Regression.

Dataset:
Choose the dataset the model will train on. Only finalized datasets appear in the list.

Dataset Version:
Select the version of the dataset to ensure compatibility between model and data.

Bands:
Displays the number of spectral bands available in the selected dataset. This is automatically set.

Dataset Processing Level:
Indicates whether the dataset uses Reflectance or Radiance data. This is inherited from the dataset and cannot be changed at the model level.

Classes:
Lists the target classes used for training. For target detection, you must specify a Target class.

Metrics Panel:​

The Metrics Panel contains two tabs:

  • Model Metrics: Displays training and validation performance (e.g., loss, accuracy, precision, recall)
  • Dataset Metrics: Provides dataset-level statistics such as class balance and spectral diversity

Initially, when the model is in Draft state, metrics will show as "No metrics available." Once training completes, this panel populates with plots and statistics for model evaluation.

Next Steps​