Spaces:
Sleeping
Sleeping
import numpy | |
class image: | |
def __init__(self, image=None, audio=None, samplerate=None, beatmap=None, log=None): | |
self.image=image | |
self.audio = audio | |
self.samplerate=samplerate | |
self.beatmap=beatmap | |
self.log=log | |
def __getitem__(self, var): | |
return self.beatmap[var] | |
def _printlog(self, string, end=None, force = False, forcei = False): | |
if (self.log is True or force is True) and forcei is False: | |
if end is None: print(string) | |
else:print(string,end=end) | |
def _toshape(self): | |
if self.image.ndim == 2: | |
self.image = [self.image] | |
if self.image.ndim == 3: | |
if len(self.image[0][0]) == 3: | |
self.image = [self.image] | |
def _channel(self): | |
if self.image.ndim == 2: yield self.image | |
if self.image.ndim == 3: | |
if len(self.image[0][0]) == 3: | |
yield self.image | |
if self.image.ndim > 3 or len(self.image[0][0] != 3): | |
for i in self.image: | |
yield i | |
def combined(self): | |
for i, channel in enumerate(self._channel()): | |
if i==0: combined = channel.copy() | |
else: combined += channel | |
return combined | |
def open(self, path): | |
import cv2 | |
self.image = cv2.imread(path) | |
def display(self): | |
import cv2 | |
cv2.imshow('image', self.combined) | |
def write(self, output, rotate = True, mode = 'square', maximum = 4096): | |
import cv2 | |
mode = mode.lower() | |
image = self.combined | |
if mode=='square': | |
y=min(len(image), len(image[0]), maximum) | |
y=max(y, maximum) | |
image = cv2.resize(image, (y,y), interpolation=cv2.INTER_NEAREST) | |
elif mode=='tosmallest': | |
y=min(len(image), len(image[0])) | |
image = cv2.resize(image, (x,x), interpolation=cv2.INTER_NEAREST) | |
elif mode=='maximum': | |
x = min(len(image), maximum) | |
y = min(len(image[0]), maximum) | |
image = cv2.resize(image, (x,y), interpolation=cv2.INTER_NEAREST) | |
if rotate is True: image=image.T | |
cv2.imwrite(output, image) | |
def effect_blur(self, value=(5,5)): | |
"""similar to echo""" | |
import cv2 | |
if isinstance(value, int) or isinstance(value, float): value = (value, value) | |
for i in range(len(self.image)): | |
self.image[i]=cv2.blur(self.image[i], value) | |
def effect_median(self, value=5): | |
"""similar to echo""" | |
import scipy.signal | |
for i in range(len(self.image)): | |
self.image[i]=scipy.signal.medfilt2d(self.image[i], value) | |
def effect_uniform(self, value=5): | |
"""similar to echo""" | |
import scipy.ndimage | |
for i in range(len(self.image)): | |
self.image[i]= scipy.ndimage.uniform_filter(self.image[i], value) | |
def effect_shift2d(self, value=5): | |
"""very weird effect, mostly produces silence""" | |
import scipy.ndimage | |
self.image= scipy.ndimage.fourier_gaussian(self.image, value) | |
self.image=self.image*(255/numpy.max(self.image)) | |
def effect_spline(self, value=3): | |
"""barely noticeable echo""" | |
import scipy.ndimage | |
for i in range(len(self.image)): | |
self.image[i]= scipy.ndimage.spline_filter(self.image[i], value) | |
def effect_rotate(self, value=0.1): | |
"""rotates self.image in degrees""" | |
import scipy.ndimage | |
image = [0 for _ in range(len(self.image))] | |
for i in range(len(image)): | |
image[i] = scipy.ndimage.rotate(self.image[i], value) | |
self.image = numpy.asarray(image) | |
def effect_gradient(self): | |
self.image=numpy.asarray(numpy.gradient(self.image)[0]) | |
class spectogram(image): | |
def generate(self, hop_length:int = 512): | |
self.hop_length=hop_length | |
import librosa | |
self.image=librosa.feature.melspectrogram(y=self.audio, sr=self.samplerate, hop_length=hop_length) | |
self.mask = numpy.full(self.image.shape, True) | |
self._toshape() | |
def toaudio(self): | |
import librosa | |
self.audio=librosa.feature.inverse.mel_to_audio(M=numpy.swapaxes(numpy.swapaxes(numpy.dstack(( self.image[0,:,:], self.image[1,:,:])), 0, 2), 1,2), sr=self.samplerate, hop_length=self.hop_length) | |
return self.audio | |
class beat_image(image): | |
def generate(self, mode='median'): | |
"""Turns song into an image based on beat positions.""" | |
assert self.beatmap is not None, 'Please run song.beatmap.generate() first. beat_image.generate needs beatmap to work.' | |
self._printlog('generating beat-image; ') | |
mode=mode.lower() | |
if isinstance(self.audio,numpy.ndarray): self.audio=numpy.ndarray.tolist(self.audio) | |
# add the bits before first beat | |
self.image=([self.audio[0][0:self.beatmap[0]],], [self.audio[1][0:self.beatmap[0]],]) | |
# maximum is needed to make the array homogeneous | |
maximum=self.beatmap[0] | |
values=[] | |
values.append(self.beatmap[0]) | |
for i in range(len(self.beatmap)-1): | |
self.image[0].append(self.audio[0][self.beatmap[i]:self.beatmap[i+1]]) | |
self.image[1].append(self.audio[1][self.beatmap[i]:self.beatmap[i+1]]) | |
maximum = max(self.beatmap[i+1]-self.beatmap[i], maximum) | |
values.append(self.beatmap[i+1]-self.beatmap[i]) | |
if 'max' in mode: norm=maximum | |
elif 'med' in mode: norm=numpy.median(values) | |
elif 'av' in mode: norm=numpy.average(values) | |
for i in range(len(self.image[0])): | |
beat_diff=int(norm-len(self.image[0][i])) | |
if beat_diff>0: | |
self.image[0][i].extend([numpy.nan]*beat_diff) | |
self.image[1][i].extend([numpy.nan]*beat_diff) | |
elif beat_diff<0: | |
self.image[0][i]=self.image[0][i][:beat_diff] | |
self.image[1][i]=self.image[1][i][:beat_diff] | |
self.image=numpy.asarray(self.image)*255 | |
self.mask = self.image == numpy.nan | |
self.image=numpy.nan_to_num(self.image) | |
self._toshape() | |
def toaudio(self): | |
self._printlog('converting beat-image to audio; ') | |
image=numpy.asarray(self.image)/255 | |
try: image[self.mask]=numpy.nan | |
except IndexError: pass | |
audio=list([] for _ in range(len(image))) | |
#print(audio) | |
#print(len(image), len(image[0]), len(image[1]), len(image[0][0]), len(image[1][0]), len(image[0][1]), len(image[1][1])) | |
for j in range(len(image)): | |
for i in range(len(image[j])): | |
beat=image[j][i] | |
#print(i,j, len(image[0][j]), len(image[1][j]), len(beat), end=' ') | |
beat=beat[~numpy.isnan(beat)] | |
#print(len(beat), end=' ') | |
audio[j].extend(beat) | |
#print(len(audio[0]), len(audio[1])) | |
self.audio=numpy.asarray(audio) | |
return self.audio |