learn.py

import sklearn 
from sklearn import linear_model
from sklearn.neighbors import KNeighborsClassifier
from sklearn.cross_validation import train_test_split
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score
import matplotlib.pyplot as plt
import scipy
import os
import sys
import glob
import numpy as np
from utils1 import GENRE_DIR, GENRE_LIST
from sklearn.externals import joblib
from random import shuffle

"""reads FFT-files and prepares X_train and y_train.
genre_list must consist of names of folders/genres consisting of the required FFT-files
base_dir must contain genre_list of directories
"""
def read_fft(genre_list, base_dir):
	X = []
	y = []
	for label, genre in enumerate(genre_list):
		# create UNIX pathnames to id FFT-files.
		genre_dir = os.path.join(base_dir, genre, "*.fft.npy")
		# get path names that math genre-dir
		file_list = glob.glob(genre_dir)
		for file in file_list:
			fft_features = np.load(file)
			X.append(fft_features)
			y.append(label)
	
	return np.array(X), np.array(y)


"""reads MFCC-files and prepares X_train and y_train.
genre_list must consist of names of folders/genres consisting of the required MFCC-files
base_dir must contain genre_list of directories
"""
def read_ceps(genre_list, base_dir):
	X= []
	y=[]
	for label, genre in enumerate(genre_list):
		for fn in glob.glob(os.path.join(base_dir, genre, "*.ceps.npy")):
			ceps = np.load(fn)
			num_ceps = len(ceps)
			X.append(np.mean(ceps[int(num_ceps*1/10):int(num_ceps*9/10)], axis=0))
			#X.append(ceps)
			y.append(label)
	
	print(np.array(X).shape)
	print(len(y))
	return np.array(X), np.array(y)

def learn_and_classify(X_train, y_train, X_test, y_test, genre_list):

	
	print(len(X_train))
	print(len(X_train[0]))

	#Logistic Regression classifier

	logistic_classifier = linear_model.logistic.LogisticRegression()
	logistic_classifier.fit(X_train, y_train)
	logistic_predictions = logistic_classifier.predict(X_test)
	logistic_accuracy = accuracy_score(y_test, logistic_predictions)
	logistic_cm = confusion_matrix(y_test, logistic_predictions)
	print("logistic accuracy = " + str(logistic_accuracy))
	print("logistic_cm:")
	print(logistic_cm)

	#change the pickle file when using another classifier eg model_mfcc_fft

	joblib.dump(logistic_classifier, 'saved_models/model_mfcc_log.pkl')

	#K-Nearest neighbour classifier

	knn_classifier = KNeighborsClassifier()
	knn_classifier.fit(X_train, y_train)
	knn_predictions = knn_classifier.predict(X_test)
	knn_accuracy = accuracy_score(y_test, knn_predictions)
	knn_cm = confusion_matrix(y_test, knn_predictions)
	print("knn accuracy = " + str(knn_accuracy))
	print("knn_cm:") 
	print(knn_cm)
	joblib.dump(knn_classifier, 'saved_models/model_mfcc_knn.pkl')
	
	plot_confusion_matrix(logistic_cm, "Confusion matrix", genre_list)
	plot_confusion_matrix(knn_cm, "Confusion matrix for FFT classification", genre_list)


def plot_confusion_matrix(cm, title, genre_list, cmap=plt.cm.Blues):
    plt.imshow(cm, interpolation='nearest', cmap=cmap)
    plt.title(title)
    plt.colorbar()
    tick_marks = np.arange(len(genre_list))
    plt.xticks(tick_marks, genre_list, rotation=45)
    plt.yticks(tick_marks, genre_list)
    plt.tight_layout()
    plt.ylabel('True label')
    plt.xlabel('Predicted label')
    plt.show()


def main():
	
	base_dir_fft  = GENRE_DIR
	base_dir_mfcc = GENRE_DIR
	
	"""list of genres (these must be folder names consisting .wav of respective genre in the base_dir)
	Change list if needed.
	"""
	genre_list = [ "blues","classical","country","disco","metal"]
	
	#genre_list = ["classical", "jazz"] IF YOU WANT TO CLASSIFY ONLY CLASSICAL AND JAZZ

	#use FFT
	# X, y = read_fft(genre_list, base_dir_fft)
	# X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = .20)
	# print('\n******USING FFT******')
	# learn_and_classify(X_train, y_train, X_test, y_test, genre_list)
	# print('*********************\n')

	#use MFCC
	X,y= read_ceps(genre_list, base_dir_mfcc)
	X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = .20)
	print("new1",X_train.shape)
	print('******USING MFCC******')
	learn_and_classify(X_train, y_train, X_test, y_test, genre_list)
	print('*********************')
	

if __name__ == "__main__":
	main()