How to visualise a dataset

You have your .json and .h5 files.

Preparing for exploration

• Start the standalone app
• In the Audio section, either
  • Fill the audio path by inputting one of your files (1)
  • Or enter the path manually (2)
• Click on the start button (3)
• The audio engine will start (4)
• Allowing you to click on the Visualisation > Explore button (5)

Standalone app default screen with audio engine OFF

Standalone app default screen with audio engine ON

info

On Windows systems, you might get a firewall notification. Accept the prompt.

Loading your dataset

The following window opens after clicking on the Explore button.

Visualisation app default screen

One more final load into the visualisation app...

Choose your .h5 and load it.

The Open panel will close and the View one will open.

This is where you select your:

• Extraction
• Band
• Integration
• Reducer

Once your selection done, the View panel automatically closes and data starts loading.

info

Depending on your dataset size, data can take minutes to load entirely.

Visualisation app View panel

Exploring

Data loading is complete!

You are now in the main visualisation mode.

This screen has three main parts:

• The scatter (1)
• The camera controls and quick actions (2)
• The panel buttons (3) reference

info

In the center, the scatter (1) displays points, either in 2D or 3D depending on your reducer setting.

Each of these points represents an audio interval, with the length of your Integration value.

The underlying represented data for each interval depends on your actual Extractor setting.

This can be either neural network embeddings that have been reduced, or a combination of them, or even with raw or acoustic indices mixed in.

Visualisation app with scatter loaded