ibm-early-programs / animal-sounds

Animal Sounds Machine Learning

Introduction to AI Machine Learning with Sound Course Material

This repo contains all the necessary code and applications to be able to complete the Introduction to AI Machine Learning with Sound course. The instructions are provided on the online course.

audio/birdsong

Contains the wav files used as test and training samples.

birdsongs

Contains the CSVs for the different types of birds used in this course.

images

Contains the images used for the Visual Recognition service. This is used in lab 5.

labs

Contains extra code to be used in the labs.

noderedflows

You can use this sample Prediction Flow to craft a node-red application that runs a prediction using input from either the microphone or a file inject

output

Contains the sound CSVs.

src

Contains the OSP Service and OSP Converter applications. Details are below.

OSP Converter

Performs Signal Processing against a directory of audio files.

This application will work against the audio files in https://www.kaggle.com/mmoreaux/audio-cats-and-dogs csv files have also been generated for the audio files in https://www.kaggle.com/rtatman/british-birdsong-dataset

Download the audio zip from kaggle and unzip to the audio directory. The application is in the src directory To install the prerequisites run

pip install -r requirements.txt  

Run the application

python ospconverter.py

The resulting csv file will be in the output folder. The Application produces a simplified table representing an audio spectogram that can be used to create machine learning models to recognize audio sounds.

An enhancement would be to take a digitial signature, eg the top 10 and bottom 10 amplitudes and the frequencies they occur at, producing a table with 41 columns per row. The first row being the class identifier. We leave that as an exercise for our consumers, though we might consider adding it ourselves in the near future.

OSP Service

Provides a web based API to performs Signal Processing against a single audio file. This application provides a web page on \/audio which can be used to test the OSP processing.

The application also provides an endpoint \/audio\/nodered suitable to be used in conjunction with a node-red file upload or microphone node.

Run the application

python ospservice.py

License

Apache License available in license

animal-sounds's Issues

importing to node-red returns error of data type

Following all steps and model is populated properly but while doing test run it returns error

Error 400 An exception occurred during Scoring with message: Value for field name COLUMN2 with datatype ScalarType(string,true) is incorrect.

CF App is unable to start need to update the code

2019-05-20T11:49:07.31+0400 [API/21] OUT Uploading bits for app with guid xxxx
2019-05-20T11:49:14.37+0400 [API/21] OUT Creating build for app with guid xxxx
2019-05-20T11:49:14.56+0400 [API/21] OUT Updated app with guid xxxxxx ({"state"=>"STARTED"})
2019-05-20T11:49:14.77+0400 [STG/0] OUT Cell xxxx creating container for instance xxxx
2019-05-20T11:49:15.72+0400 [STG/0] OUT Cell xxxxx successfully created container for instance xxxxx
27349
2019-05-20T11:49:15.91+0400 [STG/0] OUT Downloading app package...
2019-05-20T11:49:16.06+0400 [STG/0] OUT Downloaded app package (11.1K)
2019-05-20T11:49:20.37+0400 [STG/0] ERR It looks like you're deploying on a s
tack (currently set to cflinuxfs3) that's not supported by this buildpack.
2019-05-20T11:49:20.37+0400 [STG/0] ERR That could be because you're using a
recent buildpack release on a deprecated stack.
2019-05-20T11:49:20.37+0400 [STG/0] ERR If you're using the buildpack install
ed by your CF admin, please let your admin know you saw this error message.
2019-05-20T11:49:20.37+0400 [STG/0] ERR If you at one point specified a build
pack that's at git URL, please make sure you're pointed at a version that suppor
ts this stack.
2019-05-20T11:49:20.38+0400 [STG/0] ERR Failed to compile droplet: Failed to
compile droplet: exit status 44
2019-05-20T11:49:20.53+0400 [STG/0] OUT Exit status 223
2019-05-20T11:49:20.92+0400 [STG/0] OUT Cell xxxx stopping instance xxxx
2019-05-20T11:49:20.92+0400 [STG/0] OUT Cell xxxx destroying container for instance xxxxx
2019-05-20T11:49:22.06+0400 [STG/0] OUT Cell xxxxxxx successfully destroyed container for instance xxxxx
xxxxx

Prediction table not working

Hi!
I have my node-red flow configured to run a prediction in UI but the prediction is not shown in the table. I see the prediction correctly in Node-RED.
This is my flow:

csv files for birds does not exists

some csv files like finches, sky.... on lab4 does not exist