def demo_image(writer, x, n_iter):
""" Show how to add_image() to add image to tensorboard for NNabla.
:param writer: nnabla_tensorboard summary writer.
:param x: NNabla variable tensor with the format [B, C, H, W].
:param n_iter: global_step for tensorboard.
:return: nothing.
"""
tiled = tile_images(x.d) # return numpy.ndarray
writer.add_image('Image', tiled, n_iter, dataformats='HWC')
def get_tiled_image(img, channel_last=False):
assert len(img.shape) == 4
assert isinstance(img, np.ndarray)
if channel_last:
# nnabla.monitor.tile_images requests (B, C, H, W)
# (B, H, W, C) -> (B, C, H, W)
img = img.transpose(0, 3, 1, 2)
B, C, H, W = img.shape
# create tiled image. channel last image will be returned.
tiled_image = tile_images(img)
_, _, Ct = tiled_image.shape
assert C == Ct
return tiled_image
def add(self, name, var):
import nnabla as nn
from nnabla.utils.image_utils import imsave
if isinstance(var, nn.Variable):
data = var.d.copy()
elif isinstance(var, nn.NdArray):
data = var.data.copy()
else:
assert isinstance(var, np.ndarray)
data = var.copy()
assert data.ndim > 2
channels = data.shape[-3]
data = data.reshape(-1, *data.shape[-3:])
data = data[:min(data.shape[0], self.num_images)]
data = self.normalize_method(data)
if channels > 3:
data = data[:, :3]
elif channels == 2:
data = np.concatenate(
[data, np.ones((data.shape[0], 1) + data.shape[-2:])], axis=1)
tile = tile_images(data)
path = os.path.join(self.save_dir, '{}.png'.format(name))
imsave(path, tile)
def plot_stats(digits):
print("Num images:", digits.images.shape[0])
print("Image shape:", digits.images.shape[1:])
print("Labels:", digits.target[:10])
plt.imshow(tile_images(digits.images[:64, None]), **imshow_opt)
tiny_digits = import_module("nnabla/tutorial/tiny_digits")
np.random.seed(0)
imshow_opt = dict(cmap="gray", interpolation="nearest")
# ## Logistic Regression
# ### Preparing a Toy Dataset
digits = tiny_digits.load_digits(n_class=10)
tiny_digits.plot_stats(digits)
# -
data = tiny_digits.data_iterator_tiny_digits(digits,
batch_size=64,
shuffle=True)
# -
img, label = data.next()
plt.imshow(tile_images(img), **imshow_opt)
print("labels:\n", label.reshape(8, 8))
print("Label shape:", label.shape)
# ### Preparing the Computation Graph
# !Forward pass
x = nn.Variable(img.shape) # Define an image variable
with nn.parameter_scope("affine1"):
y = PF.affine(x, 10) # Output is 10 class
# -
# !Building a loss graph
t = nn.Variable(label.shape) # Define an target variable
# !Softmax Xentropy fits multi-class classification problems
loss = F.mean(F.softmax_cross_entropy(y, t))
# -
print("Printing shapes of variables")
def plot_stats(digits):
print("Num images:", digits.images.shape[0])
print("Image shape:", digits.images.shape[1:])
print("Labels:", digits.target[:10])
plt.imshow(tile_images(digits.images[:64, None]), **imshow_opt)