BriefGPT.xyz
Jul, 2023
边界加权的逻辑一致性提高了分割网络校准
Boundary-weighted logit consistency improves calibration of segmentation networks
HTML
PDF
Neerav Karani, Neel Dey, Polina Golland
TL;DR
本文提出一种基于随机转换的正则化方法，通过对训练数据中标签模糊的像素进行空间变化保持，来提高神经网络的预测准确性，实现前列腺和心脏MRI图像分割的最佳校准。
Abstract
neural network
prediction probabilities
and accuracy are often only weakly-correlated. Inherent
label ambiguity
in training data for
→