Table 1 Structure of the neural network architecture.
From: Evaluation of spatial visual perception of streets based on deep learning and spatial syntax
Framework | Layer type | Output shape | Neuronal parameters |
|---|---|---|---|
Input | Input layer | (None, 416, 416, 3) | 0 |
Encoder | Conv2D | (None, 208, 208, 64) | 9472 |
Batch normalization | (None, 208, 208, 64) | 256 | |
Activation | (None, 208, 208, 64) | 0 | |
MaxPooling2D | (None, 103, 103, 64) | 0 | |
…… |  |  | |
ZeroPadding2D | (None, 28, 28, 1024) | 0 | |
Decoder | Conv2D | (None, 26, 26, 512) | 4,719,104 |
Batch normalization | (None, 26, 26, 512) | 2048 | |
UpSampling2D | (None, 52, 52, 512) | 0 | |
ZeroPadding2D | (None, 54, 54, 512) | 0 | |
…… |  |  | |
Batch normalization | (None, 208, 208, 64) | 256 | |
Output | Softmax | (None, 43,264, 150) | 0 |
Total parameters: 14,947,414 | Â | Â | |
