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

By Seyed Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian David Reid, Silvio Savarese. 2019

5,784 citations recorded by the source corpus.

Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value. The optimal objective for a metric is the metric itself. In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss. However, IoU has a plateau making it infeasible to optimize in the case of non-overlapping bounding boxes. In this paper, we address the this weakness by introducing a generalized version of IoU as both a new loss and a new metric. By incorporating this generalized IoU ( GIoU) as a loss into the state-of-the art object detection frameworks, we show a consistent improvement on their performance using both the standard, IoU based, and new, GIoU based, performance measures on popular object detection benchmarks such as PASCAL VOC and MS COCO.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Evaluation & Benchmarks

Citations in the corpus

Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, neutral): “Furthermore, some new improvements have been made for FCOS including moving centerness to regression branch, using GIoU loss function and normalizing regression targets by corresponding strides.”
Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection (2023, methodology): “Following previous DETR-like works [2, 24, 31, 33, 58, 64], we use the L1 loss and the GIOU [41] loss for bounding box regressions.”
Learning Open-World Object Proposals without Learning to Classify (2021, neutral): “Dice coefficient [42] and generalized-IoU [43].”
DetNAS: Backbone Search for Object Detection (2019, background): “In the past few years, many advances in object detection come from the study of "head", such as architecture [13], loss [14, 23], and anchor [28, 25].”
Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection (2020, background): “We define training loss L with GFL: L = 1 Npos ∑ z LQ + 1 Npos ∑ z 1{c∗z>0} ( λ0LB + λ1LD ) , (11) where LQ is QFL and LD is DFL. Typically, LB denotes the GIoU Loss as in [26, 31].”
MMDetection: Open MMLab Detection Toolbox and Benchmark (2019, background): “Recently, there are more regression losses proposed, e.g ., Bounded IoU Loss [33], IoU Loss [32], GIoU Loss [28], Balanced L1 Loss [23].”
MDETR - Modulated Detection for End-to-End Multi-Modal Understanding (2021, methodology): “The classification head is supervised using standard cross-entropy, while the bounding box head is supervised using a combination of absolute error (L1 loss) and Generalized IoU [48].”
YOLOv4: Optimal Speed and Accuracy of Object Detection (2020, background): “For example, GIoU loss [65] is to include the shape and orientation of object in addition to the coverage area.”
DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR (2022, methodology): “L1 loss with coefficient 5.0 and GIOU loss (Rezatofighi et al., 2019) with coefficient 2.0 are consistent in both the matching and the final loss calculation procedures.”
MOTR: End-to-End Multiple-Object Tracking with TRansformer (2021, background): “Ll1 denotes L1 loss and Lgiou is the generalized IoU loss [29].”