Skip to main content

Validation Summary

The Validation Summary tab provides a centralized dashboard to track the performance of AI models across datasets. It displays key statistics, a searchable and filterable table of validation runs, and a guided workflow to start new validations.

1. Key Metrics Overview​

At the top of the page, high-level statistics offer quick insights into validation activity:

  • In Progress: Number of currently running validations
  • Completed Runs: Total number of completed validation runs
  • Model Counts: Number of distinct models used in validations
  • Dataset Counts: Number of datasets used across all runs
  • Most Used Model: The model used in the most validation runs
  • Most Used Dataset: The dataset selected most frequently for validation

These metrics help assess overall coverage, reuse, and workload distribution.

2. Validation Runs Table​

The main table lists all validation runs with the following details:

  • Validation Name: User-defined label for the run
  • Model Name and Model Type: Which model was used, and for what kind of task (e.g., segmentation)
  • Progress: Real-time status such as Completed or In Progress
  • Sample: Displayed as Total / Done / Failed for each run
  • Dataset: The dataset used, including metadata like sample source
  • Annotation: The annotation set used to compute validation metrics
  • User: The user who initiated the run
  • Begin / End Time: Timestamps for execution

3. Filters and Run Management​

Clicking the Filters button opens a side panel where you can refine the list by:

  • Model or Model Type
  • Run Status (Completed, In Progress)
  • Dataset or Annotation
  • User or Model Tags
  • Whether metrics were generated
  • Favorited status
  • Active vs. Archived runs

Filters are useful for narrowing down large-scale validations and identifying specific runs.

4. Starting a New Validation​

Click the Start Validation button to begin a new validation workflow. The guided setup includes:

4.1 Model Selection​

  • Choose from deployed models compatible with your use case.
  • Review model details such as framework, task type, and version.

4.2 Dataset Selection​

  • Select an available dataset or upload your own.
  • The system checks compatibility between dataset and model.

4.3 Annotation Selection​

  • Pick the ground truth annotation set used to evaluate model predictions.

4.4 Configure Metrics (Optional)​

  • Customize which metrics to generate, such as Dice, IoU, or F1-score.
  • Skip this step to use default metric settings.

4.5 Run Settings​

  • Add a run name and optional tags.
  • Choose to validate the full dataset or a sample subset.

4.6 Launch​

  • Click Start to trigger the run.
  • The new validation will appear in the table and update in real time.

5. Monitoring Progress​

As validations proceed, you can:

  • Track completion rates (Done vs. Total)
  • View failures instantly
  • Click into any run to review detailed metrics
  • Stop active runs with the Stop Validations button if needed

Note: The Summary tab is ideal for validation tracking at scale. Use it to explore model behavior across multiple datasets, benchmark progress, and manage validation workflows with transparency.