Variations of Squeeze and Excitation networks
This work addresses feature selection in CNNs for computer vision tasks, but it appears incremental as it builds directly on the existing SE module.
The paper tackles the problem of improving feature selection in convolutional neural networks by proposing variations of the Squeeze and Excitation (SE) module, which enhance performance as shown in experiments on residual networks.
Convolutional neural networks learns spatial features and are heavily interlinked within kernels. The SE module have broken the traditional route of neural networks passing the entire result to next layer. Instead SE only passes important features to be learned with its squeeze and excitation (SE) module. We propose variations of the SE module which improvises the process of squeeze and excitation and enhances the performance. The proposed squeezing or exciting the layer makes it possible for having a smooth transition of layer weights. These proposed variations also retain the characteristics of SE module. The experimented results are carried out on residual networks and the results are tabulated.