Models

Here we illustrate the structures of the models in our repository.

Model 1: DeepCov

Model 2: ResPRE

Model 3: Cbam-ResPRE

Model 4: SE-ResPRE model

Model 5: Halo-ResPRE

Model 6: NL-ResPRE

Model 7: FC-ResPRE for Ablation Studies

Model 8: Cbam-FC-ResPRE for Ablation Studies

Model 9: SE-FC-ResPRE for Ablation Studies

Model 10: Dilated-Resnet34 (Dilated Residual Network) for Ablation Studies

Here is a brief introduction of the dilated residual network we have implemented.

Input layer: a convolution block of output channels 16 and kernel size 7;
Input feature extractor: two convolution blocks of output channels 16 and kernel size 3, and two convolution blocks of output channels 32 and kernel size 3;
Layer 1: 3 residual blocks of output channels 64 and kernel size 3.
Layer 2: 4 residual blocks of output channels 96 and kernel size 3.
Layer 3: 6 residual blocks of output channels 128, kernel size 3 and dilation 2;
Layer 4: 3 residual blocks of output channels 160, kernel size 3 and dilation 4;
Output feature extractor: two convolution blocks of output channel 160, kernel size 3 and dilation 2, and two convolution blocks of output channel 160 and kernel size 3.
Output layer: a convolution block of output channels 10 and kernel size 1.