Table 1 Configuration of the proposed Dynamic-C2f module.
Hyperparameter | Value | Description |
|---|---|---|
Input channels | C1 | Number of channels from the previous layer |
Output channels | C2 | Number of output channels |
Number of bottlenecks (n) | 3 or 6 | Number of Bottleneck_DynamicConv blocks in the module |
Bottleneck expansion ratio (e) | 0.5 | Channel expansion ratio within each bottleneck |
Groups (g) | 1 | Groups for convolution, standard depthwise convolution is not used here |
Shortcut connection | True | Whether to use skip connections in the bottleneck |
Dynamic convolution layer | DynamicConv | Replaces the standard convolution in the second layer (cv2) |
Kernel size (k) | 3 | Kernel size for the DynamicConv layer |
Activation function | SiLU | Default activation used in the DynamicConv block |
