Abstract
Intersection over Union (IoU) is the most popular evaluation metric used inthe object detection benchmarks. However, there is a gap between optimizing thecommonly used distance losses for regressing the parameters of a bounding boxand maximizing this metric value. The optimal objective for a metric is themetric itself. In the case of axis-aligned 2D bounding boxes, it can be shownthat $IoU$ can be directly used as a regression loss. However, $IoU$ has aplateau making it infeasible to optimize in the case of non-overlappingbounding boxes. In this paper, we address the weaknesses of $IoU$ byintroducing a generalized version as both a new loss and a new metric. Byincorporating this generalized $IoU$ ($GIoU$) as a loss into the state-of-theart object detection frameworks, we show a consistent improvement on theirperformance using both the standard, $IoU$ based, and new, $GIoU$ based,performance measures on popular object detection benchmarks such as PASCAL VOCand MS COCO.