Comparison of Attention Module for Acoustic Scene Classification
Nisan Aryal, Sang-Woong Lee · 2020
Deep neural networks have seen new milestones after the introduction of attention. Attention is defined as a mechanism in deep learning in which more priority or focus is given to a certain part of the data. A different variation of attention has been introduced in recent years. In this paper, we have used Convolutional Block Attention Module and Squeeze and Excitation Networks attention module in the Resnet-18 model for acoustic scene classification. Acoustic scene classification is a variation of sound classification in which we identify the place where the sound is recorded. Our study shows that squeeze and Excitation Networks, followed by Convolutional Block Attention Module, gives 2.97% more than the baseline Resnet-18 network.