Table 4 SqueezeNet Model Architecture.
From: A deep ensemble learning approach for squamous cell classification in cervical cancer
Operation | Data Dimensions | Weights (N) | Details |
|---|---|---|---|
Input | 64x64x3 | - | Input Layer |
Conv2D | 64x64x3 \(\rightarrow\) 32x32x96 | 14208 | Convolution |
Max Pooling | 32x32x96 \(\rightarrow\) 15x15x96 | 0 | Max Pooling |
Conv2D | 15x15x96 \(\rightarrow\) 15x15x16 | 1552 | Convolution |
Conv2D | 15x15x96 \(\rightarrow\) 15x15x64 | 6208 | Convolution |
concatenate | 15x15x16 + 15x15x64 | - | Concatenation |
Conv2D | 15x15x80 \(\rightarrow\) 15x15x16 | 1296 | Convolution |
Conv2D | 15x15x80 \(\rightarrow\) 15x15x64 | 46144 | Convolution |
concatenate | 15x15x16 + 15x15x64 | - | Concatenation |
Max Pooling | 15x15x80 \(\rightarrow\) 7x7x80 | 0 | Max Pooling |
Conv2D | 7x7x80 \(\rightarrow\) 7x7x32 | 2592 | Convolution |
Conv2D | 7x7x80 \(\rightarrow\) 7x7x128 | 10368 | Convolution |
concatenate | 7x7x32 + 7x7x128 | - | Concatenation |
Conv2D | 7x7x160 \(\rightarrow\) 7x7x32 | 5152 | Convolution |
Conv2D | 7x7x160 \(\rightarrow\) 7x7x128 | 184448 | Convolution |
concatenate | 7x7x32 + 7x7x128 | - | Concatenation |
Max Pooling | 7x7x160 \(\rightarrow\) 3x3x160 | 0 | Max Pooling |
Conv2D | 3x3x160 \(\rightarrow\) 3x3x48 | 7728 | Convolution |
Conv2D | 3x3x160 \(\rightarrow\) 3x3x192 | 30912 | Convolution |
concatenate | 3x3x48 + 3x3x192 | - | Concatenation |
Conv2D | 3x3x240 \(\rightarrow\) 3x3x48 | 11568 | Convolution |
Conv2D | 3x3x240 \(\rightarrow\) 3x3x192 | 414912 | Convolution |
concatenate | 3x3x48 + 3x3x192 | - | Concatenation |
Conv2D | 3x3x240 \(\rightarrow\) 3x3x64 | 15424 | Convolution |
Conv2D | 3x3x240 \(\rightarrow\) 3x3x256 | 61696 | Convolution |
concatenate | 3x3x64 + 3x3x256 | - | Concatenation |
Conv2D | 3x3x320 \(\rightarrow\) 3x3x64 | 20544 | Convolution |
Conv2D | 3x3x320 \(\rightarrow\) 3x3x256 | 737536 | Convolution |
concatenate | 3x3x64 + 3x3x256 | - | Concatenation |
Dropout | 3x3x320 \(\rightarrow\) 3x3x320 | 0 | Dropout |
Conv2D | 3x3x320 \(\rightarrow\) 3x3x5 | 1605 | Convolution |
Pooling | 3x3x5 \(\rightarrow\) 5 | 0 | Global Avg Pooling |
FC | 5 \(\rightarrow\) 5 | 30 | Fully Connected |
