Table 2 MFR-Net V2 Network architecture
From: Coverless image steganography using morphed face recognition based on convolutional neural network
Type | Output size | |
|---|---|---|
Input |  | 224 × 224 |
Con2D 7 × 7, 64, stride 2 Batch Normalization Leaky ReLU |  | 112 × 112 |
Max pool, 3 × 3, stride 2 |  | 56 × 56 |
(Residual block, 64) × 3 |  |  |
(Residual block, 128) × 4 |  | 28 × 28 |
(Residual block, 256) × 6 |  | 14 × 14 |
(Residual block, 512) × 3 |  | 7 × 7 |
Global average pool | Â | 512 |
FC L2 norm, 300-d | FC L2 norm, 300-d | 300 |
Softmax | Â | Â |