Dataset definition
When defining your dataset this is the first important component. By defining the Annotation Taxonomy you will get access to filtered dataset metrics and a filtered dataset gallery. By doing this you can be sure that all annotations contain the same task view setup as well as the same input type, classes and properties.
Read more about the concept of annotation taxonomy under Key concepts.
To select an annotation taxonomy you locate and click it in the Annotation Taxonomy menu.
The taxonomies are sorted from A to Z. When the menu is open you can also search in the menu by writing the name of the taxonomy.
Currently only annotations with the status "Delivery Ready" are available in the Insights product. These are the completed and quality assured annotations without possible intermediate ones that got corrected later on.
Read more about annotations under Key concepts.
To further define the dataset you can decide from which input batch you want to include data. You can select one or multiple input batches.
Read more about the concept of input batches under Key concepts.
You select which input batches to include in the Input Batch menu. To select it you locate the name of the batch and click the checkbox next to it. You can select multiple batches as well.
The batches are sorted from last created to first created. When the menu is open you can also search in the menu by writing the name of the batch.
To filter the dataset in greater detail and to investigate specific scenarios you can use object and scene filters. Which of these filter types that are available for your dataset depends on its annotation taxonomy.
The filters you apply to your dataset will also be used when viewing the annotations in the annotation platform via the gallery.
Object filter Filters the dataset based on if its annotations contain or don’t contain objects with specific values.
Scene filter Filters the dataset based on if its annotations contain or don’t contain a specific scene property.
- Click either the “Create object filter” or "Create scene filter" button in the “Filter Dataset”-section.
- Select a filter type (currently only the type “Property” is available), an operator (is or is not) and one or multiple values.
- Click the "Apply Filter"-button to apply the filter to the dataset.
- The filter is added as a chip in the “Filter dataset” section.
- Click either the “Create object filter” or "Create scene filter" button in the “Filter Dataset”-section.
- Define the first condition by selecting a filter type (currently only the type “Property” is available), an operator (is or is not) and one or multiple values.
- Click the "Add another filter condition"-button.
- Define the next condition by selecting a filter type, an operator (is or is not) and one or multiple values.
- Repeat steps 3-4 if you want to add more conditions.
- When you have added all the conditions click the "Apply Filter"-button to apply the filter to the dataset.
- The filter is added as a chip in the “Filter dataset” section.
You can edit the filter by clicking the corresponding filter chip. Then the filter dialog will open again and you can adjust the filter to your liking.
You can delete a specific filter condition while you are creating a filter or when you are editing one. You delete the filter condition by clicking on the trash bin icon next to the corresponding filter condition heading, eg. "Filter condition 2".
If you want to delete a filter, just click the cancel button ("x") on the filter chip and it will be removed.
