On the summary page, you will find outliers based on object-level metrics. For each metric, we identify outliers based on its score distribution.
Box-plot outlier technique is used to identify outliers and according to its distance to the IQR, outliers are divided into two category: Severe outlier and moderate outlier.
On this page, images are ranked according to their distance to the IQR (between 25th and 75th percentiles).
Tagging feature is available on this page to tag images of interest and take action on them.