Table 2 The proposed CNN architecture.
Layer | Type | Input/Output size | Filter size/Stride | Activation | Pooling |
|---|---|---|---|---|---|
Sensor Data | Input | N×T×C | – | – | – |
Conv1 | Convolutional | N×T×F1 | 3 × 3/1 | ReLU | – |
Conv2 | Convolutional | N×T×F2 | 3 × 3/1 | ReLU | – |
Pooling1 | Max Pooling | N×T×F2 | 2 × 2 | – | 2 × 2 |
Conv3 | Convolutional | N×T×F3 | 3 × 3/1 | ReLU | – |
Conv4 | Convolutional | N×T×F4 | 3 × 3/1 | ReLU | – |
Pooling2 | Max Pooling | N×T×F4 | 2 × 2 | – | 2 × 2 |
Flatten | Flatten | N×T×F4 | – | – | – |
FC1 | Fully Connected | N×T×Hfc1 | – | ReLU | – |
FC2 | Fully Connected | N×T×Hfc2 | – | ReLU | – |
Output | Output | N×T×Output Size | – | – |  |
