TensorFlow Speech Recognition Challenge

Description:

This repository contains code for constructing an algorithm that understands simple speech commands, a neural network with EfficientNet architecture is used as an algorithm, trained on ten classes of competition data.

Motivation :

Training practice and an autonomous alternative for voice recognition, for my voice assistant for Raspberry Pi

Data:

Available here

Solution:

Train___EfficientNet.ipynb
This is a pytorch implementation neural net with architecture EfficientNet, by uploading pre-trained weights to the net, you can test it right from google colaboratory. demo.ipynb

Results:

Below is a confusion matrix of validation data.