Model Selection of Anomaly Detectors in the Absence of Labeled Validation Data
The paper presents a framework for evaluating image-based anomaly detectors using synthetically generated validation data. This method, which does not require additional training or fine-tuning, uses a small support set…
Continue reading