def render(node, settings):
"""
Given an AST node and settings, return a displayable object.
"""
tree = handle_ast(node, settings['omit_docstrings'])
with _tree_image(tree, settings) as fname:
return IPImage(filename=fname)
def compare_images(index):
original_image = Image.fromarray(get_picture_array(255 * data_ldr[index]))
new_size = (original_image.size[0] * 2, original_image.size[1])
new_im = Image.new('L', new_size)
new_im.paste(original_image, (0,0))
rec_image = Image.fromarray(get_picture_array(X_pred[index]))
new_im.paste(rec_image, (original_image.size[0],0))
new_im.save('data/test.png', format="PNG")
return IPImage('data/test.png')
def compare_images(img, img_pre):
original_image = Image.fromarray(get_picture_array(255 * img))
new_size = (original_image.size[0] * 2, original_image.size[1])
new_im = Image.new('L', new_size)
new_im.paste(original_image, (0,0))
rec_image = Image.fromarray(get_picture_array(img_pre))
new_im.paste(rec_image, (original_image.size[0],0))
new_im.save('mnist/test.png', format="PNG")
return IPImage('mnist/test.png')
def tree_image(tree):
t = Tree.fromstring(tree)
tv = SizableTreeView(t)
tv._cframe.print_to_file('.temp.ps')
im = Image.open('.temp.ps')
# trim whitespace
bg = Image.new(im.mode, im.size, im.getpixel((0, 0)))
diff = ImageChops.difference(im, bg)
diff = ImageChops.add(diff, diff, 2.0, -100)
bbox = diff.getbbox()
if bbox:
im = im.crop(bbox)
im.save('.temp.png', 'PNG')
display(IPImage(filename='.temp.png'))
for fname in ('.temp.ps', '.temp.png'):
try:
os.remove(fname)
except (IOError, WindowsError):
pass
def get_random_images():
index = np.random.randint(5000)
print index
original_image = Image.fromarray(get_picture_array(X, index))
new_size = (original_image.size[0] * 2, original_image.size[1])
new_im = Image.new('L', new_size)
new_im.paste(original_image, (0, 0))
rec_image = Image.fromarray(get_picture_array(X_pred, index))
new_im.paste(rec_image, (original_image.size[0], 0))
new_im.save('data/test.png', format="PNG")
get_random_images()
IPImage('data/test.png')
# <codecell>
## we find the encode layer from our ae, and use it to define an encoding function
encode_layer_index = map(lambda pair: pair[0], ae.layers).index('encode_layer')
encode_layer = ae.get_all_layers()[encode_layer_index]
def get_output_from_nn(last_layer, X):
indices = np.arange(128, X.shape[0], 128)
sys.stdout.flush()
# not splitting into batches can cause a memory error
X_batches = np.split(X, indices)
mu, sigma = np.mean(X.flatten()), np.std(X.flatten())
X_train = X.astype(np.float64)
X_train = (X_train - mu) / sigma
X_train = X_train.astype(np.float32)
X_out = X_train.reshape((X_train.shape[0], -1))
print("XShape: ", X.shape)
print("Making Predictions")
X_pred = (ae.predict(X_train) * sigma + mu).reshape(-1, 28,
28).astype(int).clip(
0, 255).astype('uint8')
print("Pred Shape: ", X_pred.shape)
get_random_images()
IPImage('images/orig.png')
print("Encoding Training Data")
X_encoded = ae.encode(X_train)
print("Encoded Shape: ", X_encoded.shape)
print("Decoding Data")
X_decoded = (ae.decode(X_encoded) * sigma + mu).reshape(
-1, 28, 28).astype(int).clip(0, 255).astype('uint8')
print("Decoded Shape: ", X_decoded.shape)
# check it worked:
pic_array = get_picture_array(X_decoded, np.random.randint(len(X_decoded)))
image = Image.fromarray(pic_array)
image.save('images/test.png', format="PNG")
IPImage('images/test.png')
def poulet():
# In[ ]:
#get_ipython().system(u'mkdir -p data')
#get_ipython().system(u'mkdir -p montage')
# In[ ]:
### show random inputs / outputs side by side
def get_picture_array(X, rescale=4):
array = X.reshape(256,256)
array = np.clip(array, a_min = 0, a_max = 255)
return array.repeat(rescale, axis = 0).repeat(rescale, axis = 1).astype(np.uint8())
def compare_images(index):
original_image = Image.fromarray(get_picture_array(255 * data_ldr[index]))
new_size = (original_image.size[0] * 2, original_image.size[1])
new_im = Image.new('L', new_size)
new_im.paste(original_image, (0,0))
rec_image = Image.fromarray(get_picture_array(X_pred[index]))
new_im.paste(rec_image, (original_image.size[0],0))
new_im.save('data/test.png', format="PNG")
return IPImage('data/test.png')
compare_images(2)
# compare_images(np.random.randint(50000))
# In[ ]:
## we find the encode layer from our ae, and use it to define an encoding function
def get_layer_by_name(net, name):
for i, layer in enumerate(net.get_all_layers()):
if layer.name == name:
return layer, i
return None, None
encode_layer, encode_layer_index = get_layer_by_name(ae, 'encode')
def encode_input(encode_layer, X):
return get_output(encode_layer, inputs=X).eval()
X_encoded = encode_input(encode_layer, data_ldr)
# In[ ]:
next_layer = ae.get_all_layers()[encode_layer_index + 1]
final_layer = ae.get_all_layers()[-1]
new_layer = InputLayer(shape=(None, encode_layer.num_units))
# N.B after we do this, we won't be able to use the original autoencoder , as the layers are broken up
next_layer.input_layer = new_layer
# In[ ]:
#def get_output_from_nn(last_layer, X):
# indices = np.arange(128, X.shape[0], 128)
# sys.stdout.flush()
#
# # not splitting into batches can cause a memory error
# X_batches = np.split(X, indices)
# out = []
# for count, X_batch in enumerate(X_batches):
# out.append(get_output(last_layer, X_batch).eval())
# sys.stdout.flush()
# return np.vstack(out)
def get_output_from_nn(last_layer, X):
return get_output(last_layer,X).eval()
def decode_encoded_input(X):
return get_output_from_nn(final_layer, X)
X_decoded = 256 * decode_encoded_input(X_encoded[2]) * 2
X_decoded = np.rint(X_decoded).astype(int)
X_decoded = np.clip(X_decoded, a_min = 0, a_max = 255)
X_decoded = X_decoded.astype('uint8')
print(X_decoded.shape)
pic_array = get_picture_array(X_decoded)
image = Image.fromarray(pic_array)
image.save('data/test.png', format="PNG")
IPImage('data/test.png')
# In[ ]:
enc_std = X_encoded.std(axis=0)
enc_mean = X_encoded.mean(axis=0)
enc_min = X_encoded.min(axis=0)
enc_max = X_encoded.max(axis=0)
m = X_encoded.shape[1]
# In[ ]:
n = 256
generated = np.random.normal(0, 1, (n, m)) * enc_std + enc_mean
generated = generated.astype(np.float32).clip(enc_min, enc_max)
X_decoded = decode_encoded_input(generated) * 256.
X_decoded = np.rint(X_decoded ).astype(int)
X_decoded = np.clip(X_decoded, a_min = 0, a_max = 255)
X_decoded = X_decoded.astype('uint8')
get_ipython().system(u'mkdir -p montage')
for i in range(n):
pic_array = get_picture_array(X_decoded[i], rescale=1)
image = Image.fromarray(pic_array)
image.save('montage/{0:03d}.png'.format(i), format='png')
# In[ ]:
get_ipython().system(u'montage -mode concatenate -tile 16x montage/*.png montage.png')
IPImage('montage.png')
def get_random_images():
index = np.random.randint(2017)
print index
original_image = Image.fromarray(get_picture_array(X, index))
new_size = (original_image.size[0] * 2, original_image.size[1])
new_im = Image.new('L', new_size)
new_im.paste(original_image, (0, 0))
rec_image = Image.fromarray(get_picture_array(X_pred, index))
new_im.paste(rec_image, (original_image.size[0], 0))
new_im.save('test_100.png', format="PNG")
get_random_images()
IPImage('test_100.png')
# <codecell>
## we find the encode layer from our ae, and use it to define an encoding function
encode_layer_index = map(lambda pair: pair[0], ae.layers).index('encode_layer')
encode_layer = ae.get_all_layers()[encode_layer_index]
def get_output_from_nn(last_layer, X):
indices = np.arange(128, X.shape[0], 128)
sys.stdout.flush()
# not splitting into batches can cause a memory error
X_batches = np.split(X, indices)
# información de la imagen seleccionada, incluyendo cambiar el tamaño, recortar la imagen, modificar pixeles.
# Fuentes: https://pillow.readthedocs.io/en/stable/reference/Image.html
img = Image.open(
requests.get(
'https://graffica.info/wp-content/uploads/2017/07/logo-nasa-fondo-oscuro.jpg',
stream=True).raw)
print(img)
# Para mostrar la imagen podemos usar IPython.
# Fuentes: https://ipython.readthedocs.io/en/stable/api/generated/IPython.display.html
# https://stackoverflow.com/questions/49478791/how-can-i-open-images-in-a-google-colaboratory-notebook-cell-from-uploaded-png-f
display(
IPImage(
'https://graffica.info/wp-content/uploads/2017/07/logo-nasa-fondo-oscuro.jpg'
))
# Para guardar imágenes Pillow nos proporciona el método save, que nos permite modificar la extensión del archivo.
# Fuentes: https://pillow.readthedocs.io/en/stable/reference/Image.html
# https://www.geeksforgeeks.org/python-pil-image-save-method/
img.save('test.jpg')
"""Para manejar colores y transparencias"""
# Con el módulo Image de Pillow podemos usar el método getdata para obtener y putdata para modificar la información pixel a pixel de la imagen.
# Fuente: https://pillow.readthedocs.io/en/stable/reference/Image.html#PIL.Image.Image.getdata
# Creamos un lienzo negro para modificarlo
img = Image.new('RGBA', (125, 125), (0, 0, 0, 255))
display(img)