Variations of Squeeze and Excitation networks

TL;DR

This paper introduces variations of the Squeeze and Excitation (SE) module in convolutional neural networks, improving feature selection and layer transition while maintaining SE's core characteristics, demonstrated on residual networks.

Contribution

The paper proposes novel variations of the SE module that enhance feature processing and layer transition in CNNs, building upon and improving the original SE design.

Findings

Improved performance on residual networks with proposed SE variations

Retained core characteristics of original SE module

Enhanced feature selection and layer transition

Abstract

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.