Tianchao龙虾

wing loss 损失函数笔记

Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression 论文笔记论文链接： https://arxiv.org/abs/1911.08287一、 Problem StatementBounding Box的回归分支在目标检测中很重要。目前，lnl_nln​-norm loss 对获得最优的IoU metric不是很合适，IoU Loss 只有在bounding boxes 有重叠的时候起作用，而且不会提供任何移

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection论文链接： https://arxiv.org/abs/2006.04388一、Problem Statement目前热门的One-stage detector的head末端通常会有三个representation: classification, localization, quality estima