概述：

CVPR2019 paper，主要解决物体标注时标注框模糊带来的问题，在训练时考虑到这种标注偏差，不去过分拟合标注位置（相当于一种hard样本处理方式）。paper提出一个KL Loss，来学习这种偏差，并在nms阶段使用学习到的偏差，以得到更合适的物体框。

On MS-COCO, we boost the Average Preci- sion (AP) of VGG-16 Faster R-CNN from 23.6% to 29.1%. More importantly, for ResNet-50-FPN Mask R-CNN, our method improves the AP and AP90 by 1.8% and 6.2% re- spectively, which significantly outperforms previous state- of-the-art bounding box refinement methods.

paper
github
解读

直接拿来用系列

reg.png network.png