Views¶

The Dashboard shows tabular data in the Rows panel at the bottom of the screen. What you see in this panel changes depending on what you are working on or have selected. The Dashboard provides several types of such views:

Data¶

Data views are created by opening a Run or Table. These views open actual data containers and display the data itself, not just metadata about the objects like in the navigation views.

Run¶

DoubleClick a Run row in the Runs tab will open a dedicated Run view in a tab called “Metrics”. When opening a Run you will see the three panels described in Panels, with a rows panel named Metrics.

When a 3LC Table was used as input to produce the metrics, that input data is joined and shown alongside all the metrics data for the Run. This is at the core of the 3LC workflow, allowing you to make improvements to your data based on the model predictions and metrics.

As you begin to edit your data, these revised editions are represented as new Tables. Sometimes, you might want to look at your metrics alongside the latest revision of your data with all the changes you have made, while other times you might want to look at the metrics alongside the data such as it was when the metrics were collected. To control which Table is shown together with the metrics, press 4 to open the “Details” menu, and scroll down to “Referenced Tables”. Here you can see which Tables are currently being shown, and if multiple Tables exist for the same dataset, you can switch between them.

Just like in a Table, you can modify your samples directly.

Table¶

DoubleClick a Table row in the Tables tab will open a dedicated Table view in a tab called “Examples”. When opening a Table you will see an instantiation of a particular revision of your dataset. You can modify your samples directly in this view.

Multiple Runs / Tables¶

If you want to compare the detailed results of multiple Runs or Tables, you can open them in the same session. To do that, click one Run and Ctrl + DoubleClick the other Run in the Runs page. Multiple Tables can be opened the same way.

To compare two Runs in a session, you may want to sort on the Example_id so that samples with the same ID from different Runs will be next to each other. You can create virtual columns to assist the comparison. In the figure above, a virtual column, range of predicted BBs counts for each sample, has been added. This virtual column could be filtered to >1 to see samples that have different numbers of BB predictions for the two Runs.

Warning

In order to display two Runs in the same session (i.e., all data in the same metrics table under the same METRICS tab), the two Runs must have exactly the same columns. Otherwise, the two Runs will be opened in two tabs.

You also can compare the recorded hyperparameters when you are in a multi-Run view.

Derived Views¶

Derived views are created from other data views. They allow you to work with transformed or filtered versions of your data while maintaining the original view.

Subset Table¶

A Subset table contains filtered-in data at the moment that the subset table is created. When filters and/or polygon selections are applied, you can create a subset table to work with just that portion of your dataset. Subset tables are independent from the original table and other subset tables, meaning any operations in one table will not affect others.

Reduced Table¶

A Reduced table is derived by computing averages of rows that have the same values on specified reducing column(s). This is useful for aggregating data, such as computing average metrics per epoch or creating confusion-matrix-like visualizations. Reduced tables compute averages for all applicable columns and are static once created.