Focal and Efficient IOU Loss for Accurate Bounding Box Regression

By Yi-Fan Zhang, Weiqiang Ren, Zhang Zhang, Zhen Jia, Liang Wang, Tien-Ping Tan. Neurocomputing, 2021

1,875 citations recorded by the source corpus.

In object detection, bounding box regression (BBR) is a crucial step that determines the object localization performance. However, we find that most previous loss functions for BBR have two main drawbacks: (i) Both $\ell_n$-norm and IOU-based loss functions are inefficient to depict the objective of BBR, which leads to slow convergence and inaccurate regression results. (ii) Most of the loss functions ignore the imbalance problem in BBR that the large number of anchor boxes which have small overlaps with the target boxes contribute most to the optimization of BBR. To mitigate the adverse effects caused thereby, we perform thorough studies to exploit the potential of BBR losses in this paper. Firstly, an Efficient Intersection over Union (EIOU) loss is proposed, which explicitly measures the discrepancies of three geometric factors in BBR, i.e., the overlap area, the central point and the side length. After that, we state the Effective Example Mining (EEM) problem and propose a regression version of focal loss to make the regression process focus on high-quality anchor boxes. Finally, the above two parts are combined to obtain a new loss function, namely Focal-EIOU loss. Extensive experiments on both synthetic and real datasets are performed. Notable superiorities on both the convergence speed and the localization accuracy can be achieved over other BBR losses.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/eb20c617dc1b6e5b979f830fc38af37e7f1c9d3c

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

Citations in the corpus

YOLO-FaceV2: A Scale and Occlusion Aware Face Detector (2022, methodology): “The commonly used regression loss functions in object detectors are L1/L2 loss, smooth L1 loss, IoU loss and its variants [26, 27, 28, 29].”
Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression (2021, methodology): “There are also losses designed to focus more on high IoU objects, for example, the Rectified IoU (RIoU) loss [36], and the Focal and Efficient IoU (Focal-EIoU) loss [41].”
UAV-YOLOv8: A Small-Object-Detection Model Based on Improved YOLOv8 for UAV Aerial Photography Scenarios (2023, methodology): “EIoU [37] improves based on CIoU by treating the length and width separately as penalty terms, reﬂecting the Firstly, based on the magnitude of the angle (as in Figure 3, θ and α ) between the predicted box and the real box, the predicted box rapidly moves towards the nearest axis and then…”
Wise-IoU: Bounding Box Regression Loss with Dynamic Focusing Mechanism (2023, background): “[18] argued that the irrationality of CIoU is that”
ASF-YOLO: A Novel YOLO Model with Attentional Scale Sequence Fusion for Cell Instance Segmentation (2023, methodology): “EIoU [14] minimizes the difference in width and height between the object box and anchor box, which can improve the location effect of small objects.”
Slim-neck by GSConv: a lightweight-design for real-time detector architectures (2022, neutral): “According to the definition of the CIoU, if {( w = kw gt , h = kh gt ) | k ∈ R + } , the CIoU will degenerate into the DIoU, i.e., the relative proportion of penalty terms added in the CIoU ( αv ) will not work [37].”
MPDIoU: A Loss for Efficient and Accurate Bounding Box Regression (2023, methodology): “2 and 3 show that training YOLO v7 using L MPDIoU as regression loss can considerably improve its performance compared to the existing regression losses including L GIoU , L DIoU , L CIoU , L EIoU .”
Inner-IoU: More Effective Intersection over Union Loss with Auxiliary Bounding Box (2023, methodology): “The EIoU loss function [4] is defined as follows: w c and h c are the width and height of the minimum bounding box covering target box and predicted box.”