Table 3 Autoencoder architecture for feature extraction from BreastMNIST and PneumoniaMNIST datasets
From: Private pathological assessment via machine learning and homomorphic encryption
Type | Output shape | Param # |
|---|---|---|
Conv2D | (None, 26, 26, 16) | 160 |
MaxPooling2D | (None, 13, 13, 16) | 0 |
Conv2D | (None, 13, 13, 8) | 1160 |
MaxPooling2D | (None, 7, 7, 8) | 0 |
Conv2D | (None, 4, 4, 8) | 584 |
Flatten | (None, 128) | 0 |
Reshape | (None, 4, 4, 8) | 0 |
Conv2D | (None, 4, 4, 8) | 584 |
UpSampling2D | (None, 8, 8, 8) | 0 |
Conv2D | (None, 8, 8, 8) | 584 |
UpSampling2D | (None, 16, 16, 8) | 0 |
Conv2D | (None, 28, 28, 1) | 73 |