Skip to content

Data Fusion Labeler (dFL)

The Data Fusion Labeler (dFL) is a web-based tool for interactively labeling and curating fusion datasets. Built on top of TokSearch and CMF, dFL lets researchers browse shot-by-shot diagnostics, tag interesting events (like disruptions or ELMs), and export versioned, labeled datasets for downstream analysis or machine learning.

What It’s For

  • Interactive exploration of fusion data
  • Manual tagging of disruptions, regimes, or diagnostic features
  • Creating labeled datasets for ML training
  • Curation of campaign-specific or topic-specific shot selections

dFL supports multi-signal visualizations, metadata overlays, and integration with CMF, so labels are saved as versioned, shareable artifacts. It reads the same TokSearch pipelines and cached data as the rest of the platform.

Features

  • Interactive plotting of diagnostic time series
  • Multi-shot navigation and comparison
  • Custom label definitions and schema support
  • Export to CMF repositories for traceable reuse

Coming Soon

The dFL interface will be released publicly with an upcoming platform release. It is already in use for internal curation work within the FDP project.

If you need a disruption-tagged dataset for ML, or want to collect examples of a rare operating mode, dFL is the tool for building it without leaving the platform.