The Worrisome Impact of an Inter-rater Bias on Neural Network Training

no code implementations • 12 Jun 2019 • Or Shwartzman, Harel Gazit, Ilan Shelef, Tammy Riklin-Raviv

In this paper, we highlight the issue of inter-rater bias as opposed to random inter-observer variability and demonstrate its influence on DNN training, leading to different segmentation results for the same input images.

Segmentation Semantic Segmentation