Example ML Workflow for Working with MINST

In [74]:

    

import pandas as pd
import os
import librosa

import minst.taxonomy

# Access the example notes using the notes_df
DATA_DIR = "./data/notes"
NOTES_CSV = os.path.join(DATA_DIR, "notes_index.csv")
notes_df = pd.read_csv(NOTES_CSV, index_col=0)

# Set the note_file path to the actual path
def note_fp(filename):
    return os.path.join(DATA_DIR, filename)
notes_df["note_file"] = notes_df["note_file"].map(note_fp)

# Use the taxonomy to drop any instruments that we don't care about.
notes_df = minst.taxonomy.normalize_instrument_names(notes_df).dropna()
print("Data Summary")
print(notes_df["dataset"].value_counts())
print(notes_df["instrument"].value_counts())


    

    




Data Summary
rwc             52
philharmonia    48
Name: dataset, dtype: int64
guitar         12
violin          8
clarinet        8
double-bass     8
flute           8
tuba            8
cello           8
trumpet         8
trombone        8
bassoon         8
oboe            8
horn-french     8
Name: instrument, dtype: int64

In [76]:

    

def load_mfccs_from_audio(note_file):
    y, sr = librosa.load(note_file)
    mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13).T
    # Arbitrarily averget over this now.
    return mfcc.mean(axis=0)[np.newaxis, :]

# This is going to take a minute...
mfccs = [load_mfccs_from_audio(x) for x in notes_df["note_file"]]

# put it in the dataframe for easy grabbing later
notes_df['mfcc'] = mfccs
notes_df.iloc[0]['mfcc'].shape


    

    
Out[76]:





(1, 13)

In [77]:

    

# Convert the labels
import numpy as np
import sklearn.cross_validation
import sklearn.preprocessing
enc = sklearn.preprocessing.LabelEncoder()
targets = enc.fit_transform(notes_df["instrument"])
notes_df['target'] = targets
notes_df['target']

# Make some nice SKLearn-style data
X = np.concatenate(notes_df["mfcc"])
y = np.concatenate(notes_df["target"].map(np.atleast_1d))
print(X.shape, y.shape)

# Split it into train and test
X_train, X_test, y_train, y_test = sklearn.cross_validation.train_test_split(X, y)
print(X_train.shape, y_train.shape, X_test.shape, y_test.shape)


    

    




(100, 13) (100,)
(75, 13) (75,) (25, 13) (25,)

In [78]:

    

%matplotlib inline
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestClassifier
import sklearn.metrics
clf = RandomForestClassifier(n_estimators=100)
clf.fit(X_train, y_train)

y_pred = clf.predict(X_test)
print(sklearn.metrics.classification_report(y_test, y_pred))


    

    




             precision    recall  f1-score   support

          0       0.00      0.00      0.00         2
          1       0.50      0.50      0.50         2
          2       0.00      0.00      0.00         1
          3       1.00      0.25      0.40         4
          4       0.50      0.50      0.50         2
          5       0.33      1.00      0.50         1
          6       0.00      0.00      0.00         2
          7       0.67      1.00      0.80         2
          8       0.40      0.67      0.50         3
          9       0.00      0.00      0.00         2
         10       0.67      0.67      0.67         3
         11       0.50      1.00      0.67         1

avg / total       0.45      0.44      0.39        25







    




/usr/local/lib/python3.5/site-packages/sklearn/metrics/classification.py:1074: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.
  'precision', 'predicted', average, warn_for)

In [ ]: