Table 1 AlexNet’s parameter.
From: A visual SLAM loop closure detection method based on lightweight siamese capsule network
Parameter | Convolution layer | Dense layer | ||||||
|---|---|---|---|---|---|---|---|---|
Layers | 1 | 2 | 3 | 4 | 5 | 1 | 2 | 3 |
Filters | 96 | 256 | 384 | 384 | 256 | – | – | – |
Kernel_size | 11 | 5 | 3 | 3 | 3 | – | – | – |
Stride | 4 | 1 | 1 | 1 | 1 | – | – | – |
Pool | Y | Y | N | N | Y | – | – | – |
Pool_size | 3 × 3 | 3 × 3 | – | – | 3 × 3 | – | – | – |
Dim | – | – | – | – | – | 4096 | 4096 | 1000 |
