Table 6 Structural parameters of DLoGNet
From: Towards accurate bird sound recognition through multi-scale texture-aware modeling
Layer (type) | Output shape | |
|---|---|---|
BDCM-1 | DLoG-1 | [−1, 4, 128, 128] |
Conv-1 (3*3) | [−1, 64, 128, 128] | |
BatchNorm2d-1 + ReLU-1 + MaxPool2d-1 | [−1, 64, 64, 64] | |
BDCM-2 | DLoG-2 | [−1, 256, 64, 64] |
Conv-2 (3*3) | [−1, 128, 64, 64] | |
BatchNorm2d-2 + ReLU-2 + MaxPool2d-2 | [−1, 128, 32, 32] | |
BDCM-3 | DLoG-3 | [−1, 512, 32, 32] |
Conv-3 (3*3) | [−1, 128, 32, 32] | |
BatchNorm2d-3 + ReLU-3 + MaxPool2d-3 | [−1, 128, 16, 16] | |
BDCM-4 | DLoG-4 | [−1, 512, 16, 16] |
Conv-4 (3*3) | [−1, 128, 16, 16] | |
BatchNorm2d-4 + ReLU-4 + MaxPool2d-4 | [−1, 128, 8, 8] | |
BDCM-5 | DLoG-5 | [−1, 512, 8, 8] |
Conv-5 (3*3) | [−1, 64, 8, 8] | |
BatchNorm2d-5 + ReLU-5 + MaxPool2d-5 | [−1, 64, 4, 4] | |
Fully connected layer-1 | [−1, 1024] | |
Fully connected layer-2 | [−1, 8] | |
