Computer Vision News - November 2022

58 Biomedical Image Analysis Validation Metric a Metric a Metric a Metric a Metric a Metric a Metric a Metric a Metric a Metric a           Choosing metrics for a biomedical problem is achieved in two main steps. First, the \textbf{problem fingerprint} is generated, which captures the characteristics of the driving biomedical problem relevant for metric selection. During this step, the category mapping assigns the given problem to the appropriate image processing category, namely image-level classification, semantic segmentation, object detection and instance segmentation. Finally, the \textbf{metric selection} step generates a set of suitable metrics from a category-specific pool of metrics in a domain interest-aware manner.