7. Analyze results#
Each stress test started generates its own report. Clicking on the name, you’ll be able to see all of the metrics that are generated for it. These include classification metrics such as accuracy, F1 score, precision and recall (values and curves), confusion matrices, detection based - mAP, mAR, etc, as well as more advanced information extracted from the stress test:
failure score - on the stress test overall, how much impact does the domain have on the model performance. Values closer to 0 indicate no change and are preferred. Here you can compare to the baseline failure score, which indicates the performance on the test set with no transformations.
vulnerability scores - how does the score of a particular sample change after it is being altered by the domain,
attributes - model susceptibility and model degradation depending on the different severity of the axis that were specified in the domain above. Last but not least, we leave you with an ordered list of samples that have impacted the model performance.
You can click each sample and investigate individually what is contributing to that drop in performance through our interactive tool.