Table 3 The detailed architecture of DenseNet101 and DensNet201
From: Multimodal hybrid convolutional neural network based brain tumor grade classification
No. of Layer | Size of the output | DenseNet101 | DenseNet201 |
|---|---|---|---|
Convolution block | Size of 224 × 224 | 12 × 12 with stride of 2 | 12 × 12 with stride of 2 |
Pooling_layer | Size of 112 × 112 | 6 × 6 max_pool with stride of 2 | 6 × 6 max_pool with stride of 2 |
Dense_block 1 | Size of 112 × 112 | 3 × ((conv_1 × 1), (conv_3 × 3)) | 3 × ((conv_1 × 1), (conv_3 × 3)) |
Transition_layer 1 | Size of 112 × 112 | conv_1 × 1 | conv_1 × 1 |
Size of 56 × 56 | average_pool 2 × 2 with stride of 2 | average_pool 2 × 2 with stride of 2 | |
Dense_block 2 | Size of 56 × 56 | 6 × ((conv_1 × 1), (conv_3 × 3)) | 6 × ((conv_1 × 1), (conv_3 × 3)) |
Size of 56 × 56 | conv_1 × 1 | conv_1 × 1 | |
Transition_layer 2 | Size of 56 × 56 | 7 × 7 with stride of 2 | 7 × 7 with stride of 2 |
Size of 28 × 28 | 3 × 3 max_pool with stride of 2 | 3 × 3 max_pool with stride of 2 | |
Dense_block 3 | Size of 28 × 28 | 6 × ((conv_1 × 1), (conv_3 × 3)) | 6 × ((conv_1 × 1), (conv_3 × 3)) |
Size of 28 × 28 | conv_1 × 1 | conv_1 × 1 | |
Transition_layer 3 | Size of 28 × 28 | average_pool 2 × 2 with stride of 2 | average_pool 2 × 2 with stride of 2 |
Size of 14 × 14 | 12 × ((conv_1 × 1), (conv_3 × 3) | 12 × ((conv_1 × 1), (conv_3 × 3) | |
Dense_block 4 | Size of 14 × 14 | conv_1 × 1 | conv_1 × 1 |
Size of 14 × 14 | average_pool 2 × 2 with stride of 2 | average_pool 2 × 2 with stride of 2 | |
Transition_layer 4 | Size of 14 × 14 | 24 × ((conv_1 × 1), (conv_3 × 3) | 48 × ((conv_1 × 1), (conv_3 × 3) |
Size of 7 × 7 | conv_1 × 1 | conv_1 × 1 | |
Dense_block 5 | Size of 7 × 7 | average_pool 2 × 2 with stride of 2 | average_pool 2 × 2 with stride of 2 |
Size of 7 × 7 | 16 × ((conv_1 × 1), (conv_3 × 3) | 32 × ((conv_1 × 1), (conv_3 × 3) | |
Classification_layer | Size of 1 × 1 | global_average_pool 7 × 7 | global_average_pool 7 × 7 |
1000 | Softmax layer_fully connected | Softmax layer_fully connected |