Computer Vision Metrics#
Evaluating detection models needs metrics beyond plain classification scores: predictions are boxes, and correctness is a matter of overlap. This section covers intersection over union, and revisits precision and recall in the object-detection setting.