You can use Nucleus to obtain class confusions in the form of confusion matrix to easily understand class based performance of the model and to get to the visual samples of a specific confusion easily.
Pre-reqs: Dataset, Annotations, Predictions (evaluated)
- Open dataset with annotations & predictions both uploaded
- Use the top navigation to go to the “Charts” page
- Use the drop down to select model of interest
- Scroll down to “Confusion Matrix” chart
- Hover on any cell to get a normalized number for any block
- See distribution over a specific subset by running a query e.g. metadata.weather=rain
- Click on any block to query the underlying data items in the grid view
Confusion matrix makes it very easy to get to error instances (FP, FN) in Nucleus. This unlocks powerful use cases e.g. verifying annotations, debugging models etc.
Updated over 1 year ago