dFL Walkthrough

Below is a short walkthrough of how to get started with dFL after loading your data. For more details on any section see the full documentation.

Basic dFL Design Layout

The basic layout of dFL is made up of three sections:

1. Control Panel -- Where users may adjust settings, load data, choose records, create labels, trim data, configure and run manual labeling/autolabeling, export data, etc.
2. Graph Configuration -- Where users choose graph types, signals, and various data customization and harmonizations.
3. Visualization Area -- Where graphs are plotted and manipulated, and labels are added and editted.

Loading Signals/Vectors

Once the data has been ingested using a dFL data providers, it automatically loads into the GUI web interface. The data records can be changed through the Control Panel, the signals and graph controls changed through the Graph Configuration, and the interactive graphs accessed through the Visualization Area.

Once a record is loaded, you can add a graph by selecting the "Add Graph" button above and to the right of Graph Configuration. Then, on the graph as many vectors/signals can be loaded into a 1D graph as desired.

Configuring Themes & dFL Project Layouts

Several themes are available in dFL to make data preparation more user freindly, including a selection of themes for colorblind users. To change your theme, navigate to the bottom of the Main Control Panel to the Theme dropdown.

In addition to themes, users may configure their own dFL Project Layouts. Layouts store/cache persistent user layouts, such as what graphs are loaded, what signals are loaded into each graph, what preprocessing steps are done, labels and label settings, tab names and configurations, themes, etc. dFL Project Layouts may be named, saved, exported, and loaded from the "Settings" tab in the Control Panel.

Trimming Your Data

Often one of the first things that needs to be done when analysing or processing a large dataset, is to trim your data so that the signals can be easily manipulated (and calculations performed) under the constraints of your hardware's memory. Data can be trimmed at the data ingestion level using your dFL data provider. However you can also trim your data from within the dFL GUI, and then store those settings by saving your Project Layout. Trimming can be done through the "Data Trim" dropdown in the Main Control Panel, the "Data Trim" dropdown in the "Autolabeler" tab, or by selecting a region to label and then hitting "Update Trims from Label Selection", and then updating data with trim.

By trimming you effectively reduce the loaded data in the Labeler to only the segmnet of data between the start and end points your set, reducing the data size to the trimmed region for ALL loaded data signals.

Labeling & Autolabeling

Labeling in dFL can be performed manually by activating Label Selection Mode and selecting labels from the Main Control Panel. Labels may then be drawn on the graph in the Visualization Area or manually entered by entering the start/end time in the Create/Edit/Delete Label area of the Main Control Panel. Once Labels are confirmed they may be downloaded from the View Labels tab in the Main Control Panel.

Records may also be autolabeled by selecting the Autolabeler tab in the Control Panel. There are a number of supported autolabelers in dFL, or custom autolabelers may be added by customizing the dFL Data Provider.

Harmonizing Your Data

dFL supports extensive harmonization and preprocessing pipelines for preparing your data for ML/AI training, analysis, model simulation inputs, etc. Data may be trimmed, filled, resampled, normalized, smoothed, etc., with one click in the dFL Graph Configuration panel, where all operations, order of operations, and associated metadata are stored durting export.

In addition, any custom filter/transformation may be rapidly added to the dFL GUI with simple python scripting or prompt engineering of the dFL data provider.

Exporting Your Data

Datasets and Labels may be easily exported in dFL. Labels are saved from Current Labels dropdown in the Main Control Panel, and datasets may be exported from the Export Data tab in the Control Panel.

Various export modes are supported in dFL, including Export default graphs data which exports all open graphs in a dFL Layout and any customizations loaded into every individual graph, allowing you to check/visualize your pipeline integraity before export. Alternatively, you may export your data using the Bulk Export option, which allows you to export selected records and signals using any combination of harmonization operations in dFL.

Customizing Your Workflow

dFL supports extensive workflow customization for advanced usage and workflows. For example, custom graphs can be added to the dFL GUI, custom filters/transformations can be added to the dFL GUI, and custom autolabelers can be added to the dFL GUI.