Skip to main content

Create New Model

The Create Model page allows users to begin the model registration process. Users can either:

  • Click the Create New Model button to start from scratch
  • Or continue from an existing Draft Model or Completed Model

Draft Models​

Draft models represent partially completed configurations that can be resumed later. Each draft entry shows:

  • Model Name
  • Deployment Status (e.g., Stopped)
  • Created Date
  • Modality, Problem Type, and Anatomy
  • Model Mode, Source Name, and Access (e.g., Public)

Clicking Continue resumes the draft from where you left off.

Completed Models​

Completed models are shown in a separate section and can be cloned or reused. Metadata for each model helps users choose which one to build upon.

Step 1: Model Source

Define how your model will be integrated into the platform. Available options include:

Model Source Options​

  • Import Assets: Upload from local path or URL
  • Import Docker Image: Register a Docker container
  • Register with URL: Use external hosted assets

Prediction Endpoint​

  • Must be a valid RESTful endpoint for inference
  • Click Deploy to validate the endpoint

Additional Options​

  • Proceed with deployment & verification tests
  • Idle Timeout

Side Panel​

  • Model Image: Upload a thumbnail image (max 2MB)
  • Links: Access Deployment & Prediction Test views

Click Next to continue.

Step 2: Model Details

Enter required metadata and model information:

Required Fields​

  • Model Name
  • Version
  • Class Mapping
  • Model Source

Optional Metadata​

  • Model Lineage
  • Compatible Datasets
  • Problem Type
  • Tags
  • Framework
  • Anatomy
  • Modality

Model Access Options​

  • Public
  • Grant User Permission

Click Next to proceed.

Configuration

Define runtime behavior of your model.

🛠 Runtime Model Parameters (Optional)​

You may define parameters that will be passed during inference.

Example:

{
  "threshold": 0.5
}