Fig. 12From: Effective automated pipeline for 3D reconstruction of synapses based on deep learningPrecision-recall curves of segmentation for each dataset at different pixel overlap sizes. Cut denotes the GrabCut method, growing denotes the variational region growing method. The performance of our approach is better than that of the baseline approaches of Refs. [15, 18, 19, 23]. a-c ATUM-SEM dataset at 0, 3 and 5 pixel overlap, respectively. d-f FIB-SEM dataset at 0, 3 and 5 pixel overlap, respectivelyBack to article page