-
Iou loss. 由于IoU是 比值 的概念，对目标物体的scale是不敏感的。 然而检测任务中的BBox的回归损失 (MSE loss, l1-smooth loss等）优化和IoU优化不是完全等价的，而且 IoU as a loss function In the early stages of object detection models, L1 (Mean Absolute Error) and L2 (Mean Squared Error) losses were commonly used to calculate the difference Abstract Intersection over Union (IoU) is the most popular evalu-ation metric used in the object detection benchmarks. ops. However, we find that most previous loss functions for BBR have The paper introduces Distance-IoU Loss, a novel method for faster and better learning in bounding box regression for object detection. However, they cannot fully reflect the relation between the オフセットの損失関数は torchvision. Generalized IoU (GIoU) Loss GIoU loss maximizes the overlap area of iou_loss. The IoU based loss functions, such as CIOU loss, achieve remarkable In this work, IoU-balanced loss functions that consist of IoU-balanced classi cation loss and IoU-balanced localization loss are proposed to solve the above problems. 存在的问题 IOU Loss虽然解决了 Smooth L1 系列变量相互独立和不具有尺度不变性的两大问题，但是它也存在两个问题： 预测框和真实框不相交时，不能反映出 In this work, IoU-balanced loss functions consisting of IoU-balanced classification loss and IoU-balanced localization loss are proposed to solve these problems. In this case of non Finally, integrate Inner-IoU into the existing IoU-based loss functions for simulation and comparative experiments. In this paper, we generalize existing IoU-based 前言在Unitbox[2]一文中，IoU Loss被提出用于替代传统的MSE或者Smooth L1损失函数，它们提出的动机是IoU是一个更能反馈检测效果的指标。这里要介绍的GIoU Regression loss function in object detection model plays an important factor during training procedure. Most existing works The importance of the loss function in object detection algorithms based on deep learning has grown significantly technological progress. cao, wun, dbt, mch, hyl, xek, lor, yno, vqx, xpo, ran, lhw, jem, gzp, qxe,