Data augmentation based on spatial deformations for histopathology: An evaluation in the context of glomeruli segmentation

Florian Allender, Rémi Allègre, Cédric Wemmert, Jean-Michel Dischler

Abstract: Highlights•Random deformation models compared for augmentation in U-Net glomeruli segmentation•Appropriate ranges of parameter values for each considered model are provided•Average Dice score boost by up to 0.23; similar performance for all models•Best performance increase for deformations of relatively high amplitude Framework for comparison between deformation models provided