Visualisation of Image Drift between different evaluations is not supported.
This graph illustrates the drift model's confidence in distinguishing between normal (reference) images and new (evaluated) images. A clear gap between the two curves indicates that the drift model is effectively detecting a change in the data. To illustrate what the drift model considers a "drift", the most extreme examples are shown below.
{{ $ctrl.evaluatedImageColumnName }}: Images from the evaluation dataset that the drift model identifies with the highest confidence as having drifted.
{{ $ctrl.referenceImageColumnName }}: Images from the reference dataset that the drift model identifies with the highest confidence as being normal (not drifted).
The specific visual traits that the drift model used to identify data drift can be seen by comparing the two sets of images.