Research Article
FaceFilter: Face Identification with Deep Learning and Filter Algorithm
Table 1
The structure of the neural network.
| Layer | Size in | Size out | Kernel | Feature map | No. of parameters |
| Input | 96 × 96 × 3 | | | | | ZeroPadding2D | 96 × 96 × 3 | 102 × 102 × 3 | | | | Conv1 | 102 × 102 × 3 | 48 × 48 × 64 | 7 × 7 × 3, 2 | 64 | 9472 | Norm | 48 × 48 × 64 | 48 × 48 × 64 | | | | ZeroPadding | 48 × 48 × 64 | 50 × 50 × 64 | | | | Max pool | 50 × 50 × 64 | 24 × 24 × 64 | 7 × 7 × 3, 2 | | | Conv2 | 24 × 24 × 64 | 24 × 24 × 64 | 1 × 1 × 3, 2 | 64 | 4160 | Norm | 24 × 24 × 64 | 24 × 24 × 64 | | | | ZeroPadding | 24 × 24 × 64 | 26 × 26 × 64 | | | | Conv3 | 26 × 26 × 64 | 24 × 24 × 192 | 3 × 3 × 3, 2 | 192 | 110784 | Norm | 24 × 24 × 192 | 24 × 24 × 192 | | | | ZeroPadding | 24 × 24 × 192 | 26 × 26 × 192 | | | | Max pool | 26 × 26 × 192 | 12 × 12 × 192 | 3 × 3 × 3, 2 | | | Inception 3a | 12 × 12 × 192 | 12 × 12 × 256 | | | | Inception 3b | 12 × 12 × 256 | 12 × 12 × 320 | | | | Inception 3c | 12 × 12 × 320 | 6 × 6 × 640 | | | | Inception 4a | 6 × 6 × 640 | 6 × 6 × 640 | | | | Inception 4e | 6 × 6 × 640 | 3 × 3 × 1024 | | | | Inception 5a | 3 × 3 × 1024 | 3 × 3 × 736 | | | | Average pool | 3 × 3 × 736 | 1 × 1 × 736 | 3 × 3 × 3, 1 | | | fc1 | 1 × 1 × 736 | 128 | | | | Fc | 128 | 1 | | | |
|
|