You can use Nucleus to debug your model predictions to find false positive samples by using mismatch with ground truth annotations.
Pre-reqs: Dataset, Annotations, Predictions
- Open a dataset with annotations & predictions uploaded
- Go to the objects tab and open the search options sidebar
- Under the model filter select the model you want to use
- Under the object filter select false positives i.e. (prediction & annotation mismatch)
The final results will show samples which have been assigned the wrong labels by the model. These items can be used to re-train and improve the model further.
Updated almost 2 years ago