Table 1 Final results of the experiment of selecting the optimal deep neural network architecture and the values of the training process parameters (training phase)
From: Semantic segmentation of human oocyte images using deep neural networks
No. | DNN | Acc | Loss | ||
|---|---|---|---|---|---|
T [%] | V [%] | T | V | ||
1 | DeepLab-v3-ResNet-18 (15) | 85 | 79 | 0.14 | 0.33 |
2 | DeepLab-v3-ResNet-50 (7) | 85 | 80 | 0.13 | 0.40 |
3 | DeepLab-v3-Inception-... (10) | 86 | 80 | 0.19 | 0.35 |
. | . | . | . | . | . |
7 | DeepLab-v3-Xception (8) | 81 | 80 | 0.16 | 0.34 |
. | . | . | . | . | . |
56 | fcnLayers (8) | 74 | 77 | 0.39 | 0.36 |
57 | SegNetLayers (4) | 72 | 72 | 0.64 | 0.77 |
. | . | . | . | . | . |
71 | SegNetLayers (7) | 50 | 69 | 1.82 | 0.79 |