from scipy import misc
from audioDataAnalysis.Utils import get_final_image, get_labels, get_files
import matplotlib.pyplot as plt
from numpy.random import choice
from scipy import ndimage
import numpy as np
mypath = '../KaggleData/'
image_path = '../KaggleData/spectograms/'
files = get_files(image_path)
labels = get_labels(mypath + 'train.csv', format='dict')
inv_map = {}
for k, v in labels.items():
inv_map[v] = inv_map.get(v, [])
inv_map[v].append(k)
file1 = choice(inv_map['1'], 1)[0]
file2 = choice(inv_map['0'], 1)[0]
#train10335.aiff
print('File2 printed')
im1 = misc.imread(image_path + file1[:-5] + '.png', mode='I')
im2 = misc.imread(image_path + file2[:-5] + '.png', mode='I')
im1 = get_final_image(im1, size='original', gray=True)
im2 = get_final_image(im2, size='original', gray=True)
print(im1.shape, im1.dtype)
print(im2.shape, im2.dtype)
#
# print(im.shape)
#
# data = alexnet_features.predict(im)
#
# print(data)
from Lib import aifc
from audioDataAnalysis.kagglenet import show_spec
from audioDataAnalysis.Utils import get_files
import numpy
import matplotlib.pyplot as plt
from scipy import misc
mypath = '../KaggleData/train/'
files = get_files(mypath)
file_name = numpy.random.choice(files)
with aifc.open(mypath + file_name, 'r') as f:
nframes = f.getnframes()
strsig = f.readframes(nframes)
data = numpy.fromstring(strsig, numpy.short).byteswap()
fig = plt.figure()
a = fig.add_subplot(2, 2, 1)
im = show_spec(data)
a.set_title('Original Spec')
a = fig.add_subplot(2, 2, 2)
im = show_spec(data, nfft=256, fs=2, noverlap=64)
a.set_title('Half overlap')
a = fig.add_subplot(2, 2, 3)
im = show_spec(data, nfft=256, fs=2, noverlap=128, window=numpy.blackman(256))
import os
from audioDataAnalysis.Utils import get_labels, get_files, get_final_image
import random
def get_imlist(path):
""" Returns a list of filenames for
all jpg images in a directory. """
return [os.path.join(path,f) for f in os.listdir(path) if f.endswith('.png')]
path = '../KaggleData/Spectrograms/spec_blackman/'
imlist = get_imlist(path)
labels_path = '../KaggleData/train.csv'
files_path = '../KaggleData/train/'
files = get_files(files_path)
labels = get_labels(labels_path)
whales = []
for i in files:
if labels[i] == '1':
whales.append(i)
random.shuffle(whales)
figure()
gray()
subplot(2, 5, 1)
n = 0
for i in whales[0:10]:
n += 1
from audioDataAnalysis.similarity import dtw, fastdtw, _traceback
from audioDataAnalysis.Utils import get_files
import numpy
from Lib import aifc
from time import time
if __name__ == '__main__':
# 1-D numeric
mypath = '../KaggleData/train/'
audio_files = get_files(mypath)
from sklearn.metrics.pairwise import manhattan_distances
with aifc.open(mypath + audio_files[1], 'r') as f:
nframes = f.getnframes()
strsig = f.readframes(nframes)
x = numpy.fromstring(strsig, numpy.short).byteswap()
with aifc.open(mypath + audio_files[3], 'r') as f:
nframes = f.getnframes()
strsig = f.readframes(nframes)
y = numpy.fromstring(strsig, numpy.short).byteswap()
dist_fun = manhattan_distances
print('Files opened, calculating the dtw.')
t1 = time()
dist, cost, acc, path = dtw(x, y, dist_fun)
t2 = time()
print('Enlapsed time of: ' + str(t2 - t1))
# print('Files opened, calculating the fastdtw.')
from audioDataAnalysis.Utils import get_labels, get_files
from Lib import aifc
import numpy as np
from sklearn import svm
import random
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
labels_path = '../KaggleData/train.csv'
save_path = 'results/'
path = '../KaggleData/train/'
labels = get_labels(labels_path)
files = get_files(path)
random.shuffle(files)
y = []
x = []
if __name__ == '__main__':
for i in files:
with aifc.open(path + i, 'r') as f:
nframes = f.getnframes()
strsig = f.readframes(nframes)
data = np.fromstring(strsig, np.short).byteswap()
x.append(data)
y.append(labels[i])
x = np.array(x).reshape([len(x), 4000, 1, 1])
print(x.shape)