Data Tags

Data Tags are a way to associate simple information to each dataset for classification purpose, or simply for organizing a database. The following tag types can be used:

  • true/false tags: each dataset has a binary value which can be either “true” or “false”. The default value is “false”.

  • choice tags: each dataset has one value among a user-defined list for this tag. The default value is the first one, “default”.

  • number tags: each dataset has an arbitrary number associated to this tag. The default value is 0.

The values of tags can be seen in the gallery view, in the bottom left corner of a dataset. Default values (respectively, “false”, “default” and “0”) are not shown.

_images/tags-database.png

Tip

Note that in a “choice” tag, the values are mutually exclusive, so only one choice can be active at a time. If multiple values should be active at the same time, consider using distinct true/false tags instead.

Specification

The specification of the tags (name, type, color and their order) is defined in the project wizard. Tags can be modified or deleted, and new tags can be created.

_images/tags-specification.png

Tip

Even if a tag is modified in the settings, its value on the datasets is preserved.

When creating a new tag, its type must first be chosen. Note that for ‘choice’ tags, the possible choices should be defined (these choices can later be modified). The color of a tag can be modified by clicking on its icon, and its name by double-clicking on the text.

Setting a tag value

The tag of a dataset can be changed in the side-panel in the database tab, or under the “Pointer” tool in the annotation tab.

_images/tags-setter.png

The tags of multiple datasets can be changed at the same time by selecting them in the database, opening the context menu and picking the matching tag menu.

Sorting and Filtering

The database can be sorted and filtered using tags. When sorting with a tag, the datasets with the same tag values are grouped together. The filtering is used to show only a subset of the datasets that match the specified filter. Filters can also be used in both Export and Experiments. Each tag type has its own type of filter, defined as follows:

  • true/false tags: the filter can be set to show datasets where the tag is “true”, “false” or either value.

  • choice tags: the filter can be set to any of the choices. If multiple choices are selected, datasets with any of these choices are shown.

  • number tag: the filter is a mathematical expression of equality or inequality. Filter strings are in the form of < 10.5, >= 0, == -1 or -10 <= x <= 10. All datasets whose number corresponds to the expression are shown.

_images/tags-filter.png

Multiple filters can be active at the same time. In that case, the “Match All Tags” and “Match Any Tags” part of the menu can be used to determine whether all conditions must be fulfilled, or only one of them.

Export

If the type of annotation Image Classification (see the Project Type page) is selected in the project wizard, the data tags can also be used during export to define the label encoding. A filter combination of the tags is used to generate a single output value.