def mask_image(zs, segmentation, image, background=None):
if background is None:
background = image.mean((0,1))
out = np.zeros((zs.shape[0], image.shape[0], image.shape[1], image.shape[2]))
for i in range(zs.shape[0]):
out[i,:,:,:] = image
for j in range(zs.shape[1]):
if zs[i,j] == 0:
out[i][segmentation == j,:] = background
return out
def mask_image(zs, segmentation, image, background = None): # auxiliary function, applies a mask to the given image
if(background is None):
background = image.mean((0, 1))
out = np.zeros((zs.shape[0], image.shape[0], image.shape[1], image.shape[2]))
for i in range(zs.shape[0]):
out[i, :, :, :] = image
for j in range(zs.shape[1]):
if zs[i, j] == 0:
out[i][segmentation == j, :] = background
return(out)
images_per_row = 16
# Processing the layer outputs
for layer_name, layer_activation in zip(
layer_names,
activations): #getting the layer_names and their activations
n_features = layer_activation.shape[-1] #features in the layer
size = layer_activation.shape[1] #shape of the feature map
n_cols = n_features // images_per_row #number of images per row
display_grid = np.zeros(
(size * n_cols, images_per_row * size)) #size of the display grid
for col in range(n_cols): #organizing the columns...
for row in range(images_per_row): #...and rows to display
image = layer_activation[0, :, :, col * images_per_row +
row] #retrieving the image...
image -= image.mean() #...and processing it in the...
if (image.std() > 0): #...following lines to display it
image /= image.std()
image *= 64
image += 128
image = np.clip(image, 0, 255).astype('uint8')
display_grid[col * size:(col + 1) * size,
row * size:(row + 1) * size] = image
#displaying the layer names and processed grids
print("Displaying layer:", layer_name)
scale = 1. / size
plt.figure(figsize=(scale * display_grid.shape[1],
scale * display_grid.shape[0]))
plt.title(layer_name)
plt.grid(False)