def LoadMetadata(self):
"""Loads metadata from csv files"""
plt.rcParams['figure.figsize'] = (17, 5)
# Load metadata and features.
self.tracks = utils.load("fma_metadata/tracks.csv")
self.genres = utils.load("fma_metadata/genres.csv")
self.features = utils.load("fma_metadata/features.csv")
self.echonest = utils.load("fma_metadata/echonest.csv")
np.testing.assert_array_equal(self.features.index, self.tracks.index)
assert self.echonest.index.isin(self.tracks.index).all()
return self.tracks, self.genres, self.features, self.echonest
def filter_csv():
# Load metadata and features.
tracks = utils.load('tracks.csv')
small = tracks['set', 'subset'] <= 'small'
genres = tracks.track['genre_top']
sets = tracks.set
a = pd.concat([genres, sets], axis=1)
a.to_csv('my_file.csv')
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
import librosa
import librosa.display
from collections import defaultdict
import pickle
from fma import utils
AUDIO_DIR = os.environ.get('AUDIO_DIR')
# Load metadata and features.
tracks = utils.load('fma/data/fma_metadata/tracks.csv')
genres = utils.load('fma/data/fma_metadata/genres.csv')
features = utils.load('fma/data/fma_metadata/features.csv')
echonest = utils.load('fma/data/fma_metadata/echonest.csv')
small = tracks['set', 'subset'] <= 'small'
sub = tracks.loc[small, ('track', 'genre_top')]
small_subset = echonest.loc[small]
filtered_cols = [('echonest', 'audio_features', 'acousticness'),
('echonest', 'audio_features', 'danceability'),
('echonest', 'audio_features', 'energy'),
('echonest', 'audio_features', 'instrumentalness'),
('echonest', 'audio_features', 'liveness'),
('echonest', 'audio_features', 'speechiness'),
('artist', 'active_year_end')]
for column in COLUMNS:
tracks[column] = pd.to_datetime(tracks[column])
SUBSETS = ('small', 'medium', 'large')
tracks['set', 'subset'] = tracks['set',
'subset'].astype('category',
categories=SUBSETS,
ordered=True)
COLUMNS = [('track', 'license'), ('artist', 'bio'), ('album', 'type'),
('album', 'information')]
for column in COLUMNS:
tracks[column] = tracks[column].astype('category')
return tracks
# Load metadata and features.
tracks = load(METADATA_PATH + 'tracks.csv')
#tracks = pd.read_csv(METADATA_PATH + 'tracks.csv', index_col=0, header=[0, 1])
genres = utils.load(METADATA_PATH + 'genres.csv')
features = utils.load(METADATA_PATH + 'features.csv')
echonest = utils.load(METADATA_PATH + 'echonest.csv')
# np.testing.assert_array_equal(features.index, tracks.index)
# assert echonest.index.isin(tracks.index).all()
# print('Tracks shape: ', tracks.shape, 'Generes shape: ', genres.shape,
# 'Features shape: ', features.shape, 'Echonest shape: ', echonest.shape)
print(echonest.head())
import matplotlib
matplotlib.use('TkAgg')
from matplotlib import pyplot as plt
import seaborn as sns
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
import librosa
import librosa.display
import utils
from config import rawDataPath
plt.rcParams['figure.figsize'] = (17, 5)
# Directory where mp3 are stored.
AUDIO_DIR = os.environ.get(rawDataPath)
# Load metadata and features.
tracks = utils.load('tracks.csv')
genres = utils.load('genres.csv')
features = utils.load('features.csv')
echonest = utils.load('echonest.csv')
np.testing.assert_array_equal(features.index, tracks.index)
assert echonest.index.isin(tracks.index).all()
tracks.shape, genres.shape, features.shape, echonest.shape
import IPython.display as ipd
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
from pandas import CategoricalDtype
from sklearn.metrics import recall_score
# Files
import utils
if __name__ == '__main__':
# Load metadata and features.
tracks = utils.load('src/fma_metadata/tracks.csv')
genres = utils.load('src/fma_metadata/genres.csv')
features = utils.load('src/fma_metadata/features.csv')
# assert and print shape
np.testing.assert_array_equal(features.index, tracks.index)
# print shapes
print(tracks.shape, genres.shape, features.shape)
# print heads
# ipd.display(tracks['track'].head())
# ipd.display(tracks['album'].head())
# ipd.display(tracks['artist'].head())
# ipd.display(tracks['set'].head())
import librosa
import librosa.display
import utils
import csv
# https://nbviewer.jupyter.org/github/mdeff/fma/blob/outputs/usage.ipynb
plt.rcParams['figure.figsize'] = (17, 5)
# Directory where mp3 are stored.
AUDIO_DIR = os.environ.get('/home/dan/fma_small/')
# Load metadata and features.
tracks = utils.load('tracks.csv')
features = utils.load('features.csv')
# https://www.analyticsvidhya.com/blog/2019/01/build-image-classification-model-10-minutes/
# https://colab.research.google.com/notebooks/gpu.ipynb#scrollTo=BlmQIFSLZDdc
small = tracks['set', 'subset'] <= 'small'
print(tracks['track_id'])
train = tracks['set', 'split'] == 'training'
val = tracks['set', 'split'] == 'validation'
test = tracks['set', 'split'] == 'test'
y_train = tracks.loc[small & train, ('track', 'genre_top')]
y_test = tracks.loc[small & test, ('track', 'genre_top')]
import numpy as np
import pandas as pd
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
import utils
# https://github.com/mdeff/fma/issues/9
# look at comment to get branch with scripts that work with dataset if wanted
# Load metadata and features.
tracks = utils.load('fma_metadata/tracks.csv')
# genres = utils.load('fma_metadata/genres.csv')
features = utils.load('fma_metadata/features.csv')
# echonest = utils.load('fma_metadata/echonest.csv')
# np.testing.assert_array_equal(features.index, tracks.index)
# assert echonest.index.isin(tracks.index).all()
medium = tracks['set', 'subset'] <= 'medium'
print(medium.shape)
Y = tracks.loc[medium, ('track', 'genre_top')]
X = features.loc[medium]
print(X.shape)
print('{} features, {} classes'.format(X.shape[1], np.unique(Y).size))
X.to_csv('data/dataSetMedium.csv')
Y.to_csv('data/labelsMedium.csv')
import os
import IPython.display as ipd
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
import librosa
import librosa.display
import python_utils as utils
plt.rcParams['figure.figsize'] = (17, 5)
# Directory where mp3 are stored.
AUDIO_DIR = os.environ.get(
'/Volumes/SAMSUNG/James/Physics and Machine Learning/Data/Music')
# Load metadata and features.
meta_data_path = '/Volumes/SAMSUNG/James/Physics and Machine Learning/Data/Music/fma_metadata'
tracks = utils.load(meta_data_path + '/tracks.csv')
genres = utils.load(meta_data_path + '/genres.csv')
features = utils.load(meta_data_path + '/features.csv')
echonest = utils.load(meta_data_path + '/echonest.csv')
np.testing.assert_array_equal(features.index, tracks.index)
assert echonest.index.isin(tracks.index).all()
tracks.shape, genres.shape, features.shape, echonest.shape