Skip to main content

Table 2 Final results of the experiment of selecting the optimal deep neural network architecture and the values of the training process parameters (test phase)

From: Semantic segmentation of human oocyte images using deep neural networks

No. DNN wIoU gAcc mAcc mIoU mBFS
Avg Min Max Std
1 DeepLab-v3-ResNet-18 (15) 0.897 0.93 0.79 0.97 0.03 0.74 0.62 0.80
2 DeepLab-v3-ResNet-50 (7) 0.891 0.93 0.74 0.97 0.03 0.75 0.63 0.79
3 DeepLab-v3-Inception-... (10) 0.891 0.93 0.72 0.97 0.04 0.72 0.54 0.79
. . . . . . . . . .
7 DeepLab-v3-Xception (8) 0.883 0.92 0.74 0.97 0.04 0.73 0.56 0.77
. . . . . . . . . .
56 fcnLayers (8) 0.825 0.88 0.66 0.96 0.05 0.64 0.52 0.61
57 SegNetLayers (4) 0.825 0.88 0, 65 0.96 0.06 0.49 0.36 0.66
. . . . . . . . . .
71 SegNetLayers (7) 0.749 0.82 0.58 0.95 0.08 0.32 0.22 0.50