Skip to main content

Table 2 The architecture of our deep CNN. The parameters of the input layer, convolutional layers, pooling layers, and fully connected layers are denoted as [input data size] (e.g., 64 × 64), [filter size]@[number of feature maps] (e.g., 3 × 3@32), [pooling window size] (e.g., 2 × 2), and [number of units] (e.g., 2048), respectively

From: Deep learning approach for detecting tropical cyclones and their precursors in the simulation by a cloud-resolving global nonhydrostatic atmospheric model

Layer Shape
Input 64 × 64
Convolution 1 3 × 3@32
Convolution 2 3 × 3@64
Pooling 2 × 2
Convolution 3 3 × 3@64
Pooling 2 × 2
Convolution 4 7 × 7@128
Pooling 2 × 2
Fully-connected 2048
Fully-connected 2048
Fully-connected 2