Skip to main content

Using the Playground

The Playground is an interactive environment where you can test AI models on individual images or small datasets. It supports various formats including DICOM, NIfTI, and PNG, and provides built-in tools for prediction, augmentation, and result visualization—all in real time.

1. Selecting a Dataset​

Start by choosing a dataset from the right-side panel or upload your own files via the Upload tab. You can use filters to refine your search by modality, anatomy, image format, or source. Clicking Show Details on any dataset reveals its metadata, sample count, and format type.

2. Choosing a Model​

Use the Model Type dropdown to select the kind of task (e.g., classification or segmentation). The platform automatically lists models compatible with your chosen dataset. You can inspect model specifications such as framework, version, and validation performance before applying.

3. Running Predictions​

Once a model and image are selected:

  • Use the Predict button to run inference.
  • View the result with classification labels, confidence scores, and performance metrics.
  • Advanced users can click to view raw output data in JSON format for deeper inspection.

4. Augmentation and Preprocessing​

The Playground offers image augmentation tools to simulate clinical variations:

  • Apply medical or custom augmentations like brightness, contrast, or blur.
  • Use sliders to control the intensity of transformations.
  • Restore original images with the Reset toggle.
  • You can view a summary of active augmentations and their impact on prediction outcomes.

5. Working with 3D and DICOM Images​

For volumetric datasets like CT or MRI:

  • Scroll through slices, adjust window levels, zoom, pan, and apply segmentations.
  • Access detailed DICOM metadata, including SOP tags and body parts examined.
  • Download or preview raw DICOM data directly from the interface.

6. Additional Tools​

The top toolbar includes tools for:

  • Rotating, flipping, and zooming images.
  • Drawing or editing segmentations.
  • Viewing annotation overlays.
  • Downloading prediction results and exporting metadata.

7. Quick Validation Access​

You can start validation runs directly from the Playground:

  • Use Go to Validation Metrics to review past model performance.
  • Click Start Validation to launch a new run with the current dataset and model.

8. Dataset Management​

From the Upload section, you can:

  • Upload datasets in supported formats (JPG, PNG, DICOM, NIfTI).
  • Track uploaded files, formats, and timestamps.
  • Preview dataset content and image metadata in the dataset browser.

Note: The Playground is ideal for experimenting with model performance before moving to full-scale validation workflows. Use it to test ideas quickly, fine-tune inputs, and inspect prediction quality in a controlled environment.