Faster R-CNN (Ren et al., 2015) proposes regions, then classifies and refines them. Intersection over union measures how well a proposal overlaps a ground-truth box, and it drives both training labels and the removal of duplicate detections.
Implement iou(a, b), the ratio of the overlap area to the combined area of two boxes.
[x1, y1, x2, y2].Find the intersection rectangle; if it is empty the overlap is zero. Divide the intersection area by the sum of the two areas minus the intersection.
No results yet
Run your code to check it against the examples, or submit to run every test.