Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regression

Main Idea : IoU Metric을 localization loss function으로써 사용하자.

loss = 1 - GIoU (값 : 최소 0 ~ 최대 2)

YOLO v3, Faster R CNN, Mask R CNN 네트워크에서 성능