Table 1 Network parameter structure.
From: Gaussian decomposition method for full waveform data of LiDAR base on neural network
Layers | Kernel size | Output size | Parameters |
|---|---|---|---|
Input layer | \ | \(\left(1\times 1\times 4096\right)\) | \ |
Convolution | \(\left(1\times 3, 1\right)\) | \(\left(16\times 1\times 4096\right)\) | 96 |
Dense block 1 | \(\left[\begin{array}{c}\left(1\times 5, 1\right)\\ \left(1\times 1, 1\right)\end{array}\right]\times 7\) | \(\left(128\times 1\times 4096\right)\) | 373,700 |
Transition layer | \(\left(1\times 1, 1\right)\) | \(\left(16\times 1\times 4096\right)\) | 3120 |
…… | |||
Dense block 8 | \(\left[\begin{array}{c}\left(1\times 5, 1\right)\\ \left(1\times 1, 1\right)\end{array}\right]\times 7\) | \(\left(128\times 1\times 4096\right)\) | 373,700 |
Transition layer | \(\left(1\times 1, 1\right)\) | \(\left(16\times 1\times 4096\right)\) | 3120 |
Convolution | \(\left(1\times 1, 1\right)\) | \(\left(1\times 1\times 4096\right)\) | 49 |
Output layer | \ | \(\left(1\times 4096\right)\) | \ |
