def plotstft(audiopath, generatefig=True, binsize=2**10, plotpath=None, colormap="jet"): #colormap="jet"
samplerate, samples = wav.read(audiopath)
s = stft(samples, binsize)
audio_path_split = audiopath.split('/')
if len(audio_path_split)-1 > 0:
plotpath = audio_path_split[0] + "/" + audio_path_split[1] + "/graphs/" + audio_path_split[2] # dodaj folder graphs u putanju za cuvanje grafika
plotpath = plotpath.replace('.wav', '.png') # zameni ekstenziju fajla na .png
plotpath = audiopath.replace('.wav', '.png')
sshow, freq = logscale_spec(s, factor=80.0, sr=samplerate)
ims = 20.*np.log10(np.abs(sshow)/10e-6) # amplitude to decibel
timebins, freqbins = np.shape(ims)
fig = plt.figure(figsize=(8, 4.25))
plt.imshow(np.transpose(ims), origin="lower", aspect="auto", cmap=colormap, interpolation="none")
plt.colorbar()
plt.xlabel("Time [s]")
plt.ylabel("Frequency dB[Hz]")
plt.xlim([0, timebins-1])
plt.ylim([0, freqbins])
xlocs = np.float32(np.linspace(0, timebins-1, 10))
plt.xticks(xlocs, ["%.02f" % l for l in ((xlocs*len(samples)/timebins)+(0.5*binsize))/samplerate])
ylocs = np.int16(np.round(np.linspace(0, freqbins-1, 20)))
plt.yticks(ylocs, ["%.02f" % freq[i] for i in ylocs])
#plt.clf()
fig.canvas.draw() # bitno!!! formira model grafika tj samu matricu grafika, ali je ne prikazuje korisniku!
if not(generatefig):
plt.show()
""" -temp- deo samo za prikaz sta ce ici u obucavanje mreze... posle obrisati.. """
# odlicno radi...
img_data = ImageTransform.fig2data(fig)
img_data = ImageTransform.transform(img_data)
plt.imshow(img_data, 'gray')
plt.figure()
img_data = ImageTransform.image_bin(img_data)
img_data = ImageTransform.invert(img_data)
img_data = ImageTransform.remove_noise(img_data) # zatvaranje 1.dilate 2.erode
img_data = ImageTransform.resize_graph(img_data, 70, 33) #org 350x165, 350%5=70, 165%5=33, odrzane proporcije
cv2.imwrite("test.png", img_data)
plt.imshow(img_data, 'gray')
plt.show()
else:
img_data = prepare_fig_to_img(fig) #za formiranje grafika u data-set-u ... TODO: napraviti zasebnu fun..
cv2.imwrite(plotpath, img_data)
#plt.close(fig) # sprecava memory leak - curenje memorije
return fig # vrati matlabov plot obj(numpy array)
def prepare_fig_to_img(graph_fig):
"""
@brief
Ulaz: matlabov grafik objekat
Matlabova figura postaje slika, nad slikom se vrsi
1. crop-ovanje
2. grayscale
3. binarizacija
4. uklanjanje suma
5. resize
Izlaz: slika spremna za obucavanje mreze (numpy matrica)
"""
img_data = ImageTransform.fig2data(graph_fig)
img_data = ImageTransform.transform(img_data)
img_data = ImageTransform.image_bin(img_data)
img_data = ImageTransform.invert(img_data)
img_data = ImageTransform.remove_noise(img_data) # zatvaranje 1.dilate 2.erode
img_data = ImageTransform.resize_graph(img_data, 70, 33) #org 350x165, 350%5=70, 165%5=33, odrzane proporcije
return img_data
def prepare_fig_to_img(graph_fig):
"""
@brief
Ulaz: matlabov grafik objekat
Matlabova figura postaje slika, nad slikom se vrsi
1. crop-ovanje
2. grayscale
3. binarizacija
4. uklanjanje suma
5. resize
Izlaz: slika spremna za obucavanje mreze (numpy matrica)
"""
img_data = ImageTransform.fig2data(graph_fig)
img_data = ImageTransform.transform(img_data)
img_data = ImageTransform.image_bin(img_data)
img_data = ImageTransform.invert(img_data)
img_data = ImageTransform.remove_noise(
img_data) # zatvaranje 1.dilate 2.erode
img_data = ImageTransform.resize_graph(
img_data, 70, 33) #org 350x165, 350%5=70, 165%5=33, odrzane proporcije
return img_data
img_mean = (0.4914, 0.4822, 0.4465)
img_std = (0.2023, 0.1994, 0.2010)
# Transforms
transform_train = [
transforms.RandomCrop(32, padding=4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(img_mean, img_std),
]
transform_test = [
transforms.ToTensor(),
transforms.Normalize(img_mean,img_std),
]
t_transform_train = ImageTransform.transform(transform_train)
t_transform_test = ImageTransform.transform(transform_test)
# Dataset and DataLoader arguments
dataset_name = torchvision.datasets.CIFAR10
trainSet_dict = dict(root='./data', train=True, download=True, transform=t_transform_train)
trainLoad_dict = dict(batch_size=32, shuffle=True, num_workers=4)
testSet_dict = dict(root='./data', train=False, download=True, transform=t_transform_test)
testLoad_dict = dict(batch_size=32, shuffle=False, num_workers=4)
IMAGE_PATH = "images/"
MODEL_PATH = "model/"
def main():
# Device
SEED = 1
cuda = torch.cuda.is_available()
def plotstft(audiopath,
generatefig=True,
binsize=2**10,
plotpath=None,
colormap="jet"): #colormap="jet"
samplerate, samples = wav.read(audiopath)
s = stft(samples, binsize)
audio_path_split = audiopath.split('/')
if len(audio_path_split) - 1 > 0:
plotpath = audio_path_split[0] + "/" + audio_path_split[
1] + "/graphs/" + audio_path_split[
2] # dodaj folder graphs u putanju za cuvanje grafika
plotpath = plotpath.replace('.wav',
'.png') # zameni ekstenziju fajla na .png
plotpath = audiopath.replace('.wav', '.png')
sshow, freq = logscale_spec(s, factor=80.0, sr=samplerate)
ims = 20. * np.log10(np.abs(sshow) / 10e-6) # amplitude to decibel
timebins, freqbins = np.shape(ims)
fig = plt.figure(figsize=(8, 4.25))
plt.imshow(np.transpose(ims),
origin="lower",
aspect="auto",
cmap=colormap,
interpolation="none")
plt.colorbar()
plt.xlabel("Time [s]")
plt.ylabel("Frequency dB[Hz]")
plt.xlim([0, timebins - 1])
plt.ylim([0, freqbins])
xlocs = np.float32(np.linspace(0, timebins - 1, 10))
plt.xticks(xlocs, [
"%.02f" % l for l in ((xlocs * len(samples) / timebins) +
(0.5 * binsize)) / samplerate
])
ylocs = np.int16(np.round(np.linspace(0, freqbins - 1, 20)))
plt.yticks(ylocs, ["%.02f" % freq[i] for i in ylocs])
#plt.clf()
fig.canvas.draw(
) # bitno!!! formira model grafika tj samu matricu grafika, ali je ne prikazuje korisniku!
if not (generatefig):
plt.show()
""" -temp- deo samo za prikaz sta ce ici u obucavanje mreze... posle obrisati.. """
# odlicno radi...
img_data = ImageTransform.fig2data(fig)
img_data = ImageTransform.transform(img_data)
plt.imshow(img_data, 'gray')
plt.figure()
img_data = ImageTransform.image_bin(img_data)
img_data = ImageTransform.invert(img_data)
img_data = ImageTransform.remove_noise(
img_data) # zatvaranje 1.dilate 2.erode
img_data = ImageTransform.resize_graph(
img_data, 70,
33) #org 350x165, 350%5=70, 165%5=33, odrzane proporcije
cv2.imwrite("test.png", img_data)
plt.imshow(img_data, 'gray')
plt.show()
else:
img_data = prepare_fig_to_img(
fig
) #za formiranje grafika u data-set-u ... TODO: napraviti zasebnu fun..
cv2.imwrite(plotpath, img_data)
#plt.close(fig) # sprecava memory leak - curenje memorije
return fig # vrati matlabov plot obj(numpy array)
