Atlas was developed to allow labeling of multimodal data from experiments (i.e. video and audio), but also other physical input types i.e. heart rate information.
The main aim of Atlas is to provide a graphical representation of input data and to allow the user to label provided data with attributes. Additionally Atlas allows auto-labeling by using existing labels and searching for matches in the feature input. Features used for such auto-labeling can either be raw input data from the experiment or data derived from such raw data.
All input data in Atlas is considered a track. The structure of track types is shown in the image below:

Tracks can be divided into media-tracks (video and audio) and line-tracks (also called data-tracks). Video-tracks are displayed using the VLC media player. Audio-tracks also use VLC and should be supplied in WAV format which enables Atlas to create a scalar-track as visual representation of the audio input (without the audio's scalar-track Atlas can only use the audio-track for playback).
Line-tracks are shown in Atlas' LabelArea, an area for visual represention of input data. Scalar-tracks are two-dimensional (value at a specific time) input formats like audio amplitude. Vector-tracks are multidimensional graphical formats (vector at a specific time). Vector-tracks are often used to import more complex data that is created from the experiment's original output.
Label-tracks are user defined tracks to mark wanted properties in the input tracks. For example a label-track might show a test subjects emotional state. The user can create so called label-classes with respective entities in order to label certain monitored characteristics. For example after creating a label-class 'emotional state', the user can create classes' entities (i.e. 'happy' or 'sad') and label the input data in respect to the state of the test subject.
In auto-labeling mode the user assigns a feature-track to a label-class. Atlas will then attempt to automatically create labels for the label-track with the supplied feature-track as input reference.
The combination of all input types and LabelTracks is called 'collection'.
Table 1.1. Track overview
| Track | Use |
|---|---|
| Video | Video tracks are played back using the VLC media player. VLC supports playback of many video codec formats. |
| Audio | Audio-track playback is supported for WAV and mp3 formats. When using WAV as input, Atlas can generate a scalar-track of the input file. Wav input is therefore preferable. Please note that it is possible to transcode other audio formats to wav with many audio tools. |
| Scalar | Two dimensional input (value at a specific time). Scalar-tracks are optionally created when importing wav files. |
| Vector | Multidimensional input (vector at a specific time). Used for more complex input types. |
| Label | User defined track to allow labeling of criteria the user is interested in. Each Label-track uses a label-class representing a characteristic that is monitored. In the label-track labels with entities (different predefined states of the monitored characteristic) are used. |