Table 11 The layer-wise summary of the 2D (left) and 3D (right) convolutional autoencoders used in experiments. w is the input spectral dimension, and r is the desired reduced spectral dimension size. Note that layer 1F in each network is the end of the encoder and 2A is the start of the decoder
From: CEU-Net: ensemble semantic segmentation of hyperspectral images using clustering
Layer # | Layer Name | Output Shape | Layer # | Layer Name | Output Shape |
|---|---|---|---|---|---|
0 | Input Layer | (1,1,w) | 0 | Input Layer | (1,1,1,w) |
1A | Conv2D_1 | (1,1,w) | 1A | Conv3D_1 | (1,1,1,w) |
1B | MaxPooling2D_1 | (1,1,w) | 1B | MaxPooling3D_1 | (1,1,1,w) |
1C | Conv2D_2 | (1,1,60) | 1C | Conv3D_2 | (1,1,1,60) |
1D | MaxPooling2D_2 | (1,1,60) | 1D | MaxPooling3D_2 | (1,1,1,60) |
1E | Conv2D_3 | (1,1,r) | 1E | Conv3D_3 | (1,1,1,r) |
1F | MaxPooling2D_3 | (1,1,r) | 1F | MaxPooling3D_3 | (1,1,1,r) |
2A | Conv2D_4 | (1,1,r) | 2A | Conv3D_4 | (1,1,1,r) |
2B | UpSampling2D_1 | (1,1,r) | 2B | UpSampling3D_1 | (1,1,1,r) |
2C | Conv2D_5 | (1,1,60) | 2C | Conv3D_5 | (1,1,1,60) |
2D | UpSampling2D_2 | (1,1,60) | 2D | UpSampling3D_2 | (1,1,1,60) |
2E | Conv2D_6 | (1,1,w) | 2E | Conv3D_6 | (1,1,1,w) |
2F | UpSampling2D_3 | (1,1,w) | 2F | UpSampling3D_3 | (1,1,1,w) |
2G | Conv2D_7 | (1,1,w) | 2G | Conv3D_7 | (1,1,1,w) |