/BC-ResNets

Broadcasting-residual network model zoo for Multiple Classification Tasks.

Primary LanguagePythonMIT LicenseMIT

BC-ResNets

Broadcasting-residual network model zoo for Multiple Classification Tasks.

Small but Effective

unofficial implementation of 'Broadcasted Residual Learning for Efficient Keyword Spotting'

How to Use

1. add this python file to your model directory

2. import the 'BCResNet' object and ready to go!

HyperParams Configurations

num_classes: total classification classes based on your task

delta: channel scaling, refers to τ in the original paper

aug: choose data augmentation or not

Model Params List

BC-ResNet-1: 9.2k

BC-ResNet-2: 27.3k

BC-ResNet-3: 54.2k

BC-ResNet-6: 188k

BC-ResNet-8: 321k