With this workflow you can obtain class distributions on ground truth and model predictions. This can be very useful if you want to have an informed approach to future data collection, annotation or model improvement.
Pre-reqs: Dataset, Annotations, Predictions
- Open dataset with either annotations or predictions or both uploaded
- Use the top navigation to go to the “Charts” page
- Use the drop down to select Ground Truth or model of interest
- Scroll down to the “Object Class Distribution” chart
- Hover on any bar to get an absolute number for any class
- See distribution over a specific subset by running a query e.g. metadata.weather=rain
- Click on any bar to query the underlying data items in the grid view
The chart is always adjusted based on the query. Therefore you can get class distributions over the exact subset of data you are interested in. In case you have curated slices in Nucleus, you can also get distribution on that either by running a slice name query or by going to insights page when viewing a slice.
Updated about 1 year ago