Table 1 Output shape and number of parameters for Spatial-autoencoder and FFT-autoencoders for HAM10000 and ISIC-2017 dataset.
Layer name | Output shape (HAM10000) | Output shape (ISIC-2017) | #Params | |
|---|---|---|---|---|
Encoder | Input | [28,28,3] | [64,64,3] | 0 |
Conv1 | [28,28,32] | [64,64,32] | 896 | |
Channel_attention1 | [28,28,32] | [64,64,32] | 256 | |
Maxpool1 | [14,14,32] | [32,32,32] | 0 | |
Conv2 | [14,14,16] | [32,32,16] | 4624 | |
Channel_attention1 | [14,14,16] | [32,32,16] | 64 | |
Maxpool2 | [7,7,16] | [16,16,16] | 0 | |
Decoder | Conv3 | [7,7,16] | [16,16,16] | 2320 |
UpSample1 | [14,14,16] | [32,32,16] | 0 | |
Conv4 | [14,14,32] | [32,32,32] | 4640 | |
UpSample2 | [28,28,32] | [64,42,32] | 0 | |
Conv5 | [28,28,3] | [64,64,3] | 867 | |
Total parameters | 13,667 |
