Giters

ZJULearning / RMI

This is the code for the NeurIPS 2019 paper Region Mutual Information Loss for Semantic Segmentation.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

The intuition behind RMI loss

ydhongHIT opened this issue · comments

commented

Since the intuition behind RMI loss is to model the dependencies among pixels, the improvement of boundary segmentation should be not obvious. Is it right?

commented

Well, RMI has the ability to capture the local pattern. Possibly the loss makes the pattern of the boundaries of prediction and GT more similar.

commented

Well, RMI has the ability to capture the local pattern. Possibly the loss makes the pattern of the boundaries of prediction and GT more similar.

Thanks for your reply.