Abstract:
Deep learning for pathological examination is a booming trend in the current times.
Many studies have been conducted to solve pathological problems using deep learning.
However, since the properties of an acquired image vary with the acquisition equipment, a
deep learning model would fail if the variance of an unseen image is different from that of the
images in the training data. This study aims to find a method that transfers the training set
variance on unseen images and helps to make the model inference more robust.
