Skip to content

Collection of commercial music recordings distributed in the Kenyan coast between 1930-1975

License

Notifications You must be signed in to change notification settings

chrispla/Eisenberg_Collection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 

Repository files navigation

Eisenberg Collection

Introduction

The Eisenberg Collection is a collection of commercial music recordings distributed in the Kenyan coast between 1930-1975.

Metadata

Metadata are provided as a pandas DataFrame, which is provided in HDF5 file format as a PyTables Table structure. The metadata.h5 file can be loaded using the pandas.read_hdf method.

import pandas as pd
metadata = pd.read_hdf("metadata.h5")

Not all metadata are available for all recordings, and this is the purpose of specific datasets, as explained later in this document. Empty entries contain the keyword NA. The metatags are listed below.

  • Song Title
  • Artist
  • Composer
  • Year (estimate/range)
  • Original Form
  • Label
  • Serial Number
  • Genre
  • Collection ID
  • NYUAD Archives Reference

Features

Features are provided as a python dictionary using the NYUAD Archive Reference of each recording as a key. The features.h5 file can be loaded using a library such as deepdish with the io.load method.

import deepdish as dd
features = dd.io.load("features.h5")

The archive reference for each recording is present in the metadata file, and can also be useful for retrieving the audio files of the collection if the user has access to them. The value of each key is a dictionary of features. The string keys for the features provided are

All features were computed using librosa with the default parameters for each method.

Below is an example of the features dictionary.

{
  "AD-MC-035_ref2": {  # NYUAD Archive reference key
    "mfcc": np.ndarray,  # shape=(20, t)
    "chromagram": np.ndarray,  # shape=(12, t)
    "tempogram": np.ndarray  # shape=(384, n)
   }
   ..
}

Datasets

Datasets containing subsets of the collection are provided to ensure the completness of the relevant metadata for particular tasks. For each dataset, a .txt file with newline-separated entries is provided containing the NYUAD Archive reference keys of the recordings in the dataset. The file can be loaded as a python list and used to retrieve the relevant metadata and features.

Tarab Classification

Genre Classification

Instrument Classification

Date Estimation

About

Collection of commercial music recordings distributed in the Kenyan coast between 1930-1975

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published