csharp-deep-learning-audio

Deep Learning for Audio in CSharp and Keras

Environment Configuration Before Install

The packages runs in x64 and built with .NET 4.6.1. Therefore you need to

• Set the "Target framework" to ".NET Framework 4.6.1" in the Application tab of your project Properties
• Set the "Platform Target" to x64 in the Build tab of your project Properties
• Set the .NET configuration manager to x64 in Visual Studio IDE

Install

Make sure that you have the following installed:

• TensorFlowSharp Version 1.7.0 (used to run the tensorflow audio classifier trained models)

Run the following command in your nuget manager console:

Install-Package TensorFlow-Deep-Music -Version 1.0.3

The following dlls are also installed when installing the TensorFlow-Deep-Music:

• Bass.NET Version 1.0.0 (used to convert audio file to melgram)
• System.ValueTuple Version 4.4.0 (required by TensorFlowSharp)

Usage

Audio Encoder and Audio Classifier

Below is the code showing how to use AudioClassifier to predict the genres of an audio file:

using System;
using csharp_deep_learning_audio;
using System.IO;

namespace csharp_deep_learning_audio_samples
{
    class Program
    {
        static void Main(string[] args)
        {
            AudioClassifier c = new AudioClassifier();
            string dataDirPath = @"C:\Users\chen0\git\csharp-deep-learning-audio\gtzan\genres";

            string[] subDirectories = Directory.GetDirectories(dataDirPath);
            foreach(string subDirectory in subDirectories)
            {
                string[] files = Directory.GetFiles(subDirectory, "*.au");
                foreach(string file in files)
                {
                    Console.WriteLine("classifing: {0}", file);
                    Console.WriteLine("predicted: {0}", c.PredictLabel(file));
                    break;
                }
                
            }
        }
    }
}

Audio Search Engine

Below is the code showing how to index music files and perform query to find similar songs:

using System;
using csharp_deep_learning_audio;
using System.IO;
using System.Collections.Generic;

namespace csharp_deep_learning_audio_samples
{
    class SearchEngineDemo
    {
        public static void Demo()
        {
            
            AudioSearchEngine c = new AudioSearchEngine();
            string dataDirPath = Path.Combine(IOUtils.AssemblyDirectory, "..", "..", "..", "..", "gtzan", "genres");

            string[] subDirectories = Directory.GetDirectories(dataDirPath);
            foreach(string subDirectory in subDirectories)
            {
                string[] files = Directory.GetFiles(subDirectory, "*.au");
                foreach(string file in files)
                {
                    c.IndexAudio(file);
                }
            }

            int topK = 10;
            List<ComparableAudioEntry> result = c.Query(
                Path.Combine(IOUtils.AssemblyDirectory, "..", "..", "..", "..", "music-samples", "example.mp3"),
                topK);

            for(int i=0; i < result.Count; ++i)
            {
                Console.WriteLine("#{0} (Score: {1}): {2}", i + 1, result[i].Distance, result[i].AudioFile);
            }




        }
    }
}