Deep Trainer / U-Net Bone Segmentation 020719



University of Michigan

Daniella Patton

pattondmumich.edu

Architecture: Unet
Training: 6 slices were extracted from the femoral neck of 21 different bones (Male: n=11, 24-89 years, Female: n=19, 27-95 years) for a total of 126 slices
Training Parameters: Epoch Numer: 150, Stride to Input: 1, Patch Size: 176, Batch Size: 4
Training Set: 114
Validation Set: 12
Results: Epoch 150/150
- lr: 0.0000000001 - loss: 0.0138 - acc: 0.9945 - val_loss: 0.0201 - val_acc: 0.9925



Installing extensions requires ORS Dragonfly 3.1 or later.