keras backend placeholder as input ERROR

[leoliuf]

Hello,

I tried to run the sample viz.py code with Theano as backend and got the error message as

Using Theano backend.
Namespace(img=None, iterations=5, layer='conv5_1', num_filters=16, size=128, weights_path='../vgg16_weights.h5')
Traceback (most recent call last):
  File "/usr/local/pycharm/helpers/pydev/pydevd.py", line 1580, in <module>
    globals = debugger.run(setup['file'], None, None, is_module)
  File "/usr/local/pycharm/helpers/pydev/pydevd.py", line 964, in run
    pydev_imports.execfile(file, globals, locals)  # execute the script
  File "/home/fliu/data2/data/nist/keras-filter-visualization/viz.py", line 87, in <module>
    first_layer.input = input_placeholder
AttributeError: can't set attribute

The source code snippet in viz.py around line 87 is

input_placeholder = K.placeholder((1, 3, img_width, img_height))
first_layer = ZeroPadding2D((1, 1), input_shape=(3, img_width, img_height))
first_layer.input = input_placeholder
model = get_model(first_layer)
model = load_model_weights(model, args.weights_path)
layer = get_output_layer(model, args.layer)

What does the first_layer.input do here? I am using Keras 1.2.0 and Theano 0.8.2

Thanks

[AnushaManila]

Hello,

I have the same issue (error: can't set attribute in line 87) unsolved.

I am using TensorFlow as backend. Any help would be appreciated.

Thank you.

[asampat3090]

I'm not sure how this library works but it seems to be based on an older implementation of this code. there's a great blog post about how this works (though you'll want to look at the code to figure out how to make it work with the latest keras version). If you use the code referenced you should be able to get filters and modify the code from there.

[rcasiodu]

@leoliuf @AnushaManila I have modified to keras 2.0 and add tensorflow backend.

[beatriz-ferreira]

Hi,

I am facing exactly the same error. I am using Keras 2.1.4 with Tensorflow backend. Tried to downgrade to keras 2.0 (which seemed to be the suggestion of @rcasiodu ) but got the same error... Any update on this issue?

Thank you in advance!

[rcasiodu]

@beatriz-ferreira I use python 3.6.4 and keras 2.1.3; try the following commands and modify the weight file path at the beginning.

from keras import backend as K
from keras.models import Sequential
from keras.layers import Conv2D, ZeroPadding2D, MaxPooling2D
from keras.layers.core import Flatten, Dense, Dropout

import numpy as np
import cv2

from tqdm import tqdm

#Configuration:
img_width, img_height = 128, 128
input_shape = (img_width, img_height, 3)
num_filters = 16
iterations = 20
weights_path = '../.keras/models/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5'
layer_name = 'conv5_1'
img_path = None
filter_indexes = range(0, num_filters)

def save_filters(filters, img_width, img_height):
    margin = 5
    n = int(len(filters)**0.5)
    width = n * img_width + (n - 1) * margin
    height = n * img_height + (n - 1) * margin
    stitched_filters = np.zeros((width, height, 3))

    # fill the picture with our saved filters
    for i in range(n):
        for j in range(n):
            index = i * n + j
            if index < len(filters):
                img = filters[i * n + j]
                stitched_filters[(img_width + margin) * i: (img_width + margin) * i + img_width,
                                 (img_height + margin) * j: (img_height + margin) * j + img_height, :] = img

    # save the result to disk
    cv2.imwrite('stitched_filters_%dx%d.png' % (n, n), stitched_filters)

# util function to convert a tensor into a valid image
def deprocess_image(x):
    # normalize tensor: center on 0., ensure std is 0.1
    x -= x.mean()
    x /= (x.std() + 1e-5)
    x *= 0.1

    # clip to [0, 1]
    x += 0.5
    x = np.clip(x, 0, 1)

    # convert to RGB array
    x *= 255
    
    x = np.clip(x, 0, 255).astype('uint8')
    return x

# vgg16 without 3 fully connected layer
def get_model(input_shape):
    model = Sequential()
    model.add(ZeroPadding2D((1, 1), input_shape=input_shape))
    model.add(Conv2D(64, (3, 3), activation='relu', name='conv1_1'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(64, (3, 3), activation='relu', name='conv1_2'))
    model.add(MaxPooling2D((2, 2), strides=(2, 2)))

    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(128, (3, 3), activation='relu', name='conv2_1'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(128, (3, 3), activation='relu', name='conv2_2'))
    model.add(MaxPooling2D((2, 2), strides=(2, 2)))

    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(256, (3, 3), activation='relu', name='conv3_1'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(256, (3, 3), activation='relu', name='conv3_2'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(256, (3, 3), activation='relu', name='conv3_3'))
    model.add(MaxPooling2D((2, 2), strides=(2, 2)))

    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv4_1'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv4_2'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv4_3'))
    model.add(MaxPooling2D((2, 2), strides=(2, 2)))

    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv5_1'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv5_2'))
    model.add(ZeroPadding2D((1, 1)))
    model.add(Conv2D(512, (3, 3), activation='relu', name='conv5_3'))
    model.add(MaxPooling2D((2,2), strides=(2,2)))
    return model

def get_output_layer(model, layer_name):
    # get the symbolic outputs of each "key" layer (we gave them unique names).
    #layer_dict = dict([(layer.name, layer) for layer in model.layers])
    #layer_output = layer_dict[layer_name].output
    layer_output = model.get_layer(layer_name).output
    return layer_output

def normalize(x):
    # utility function to normalize a tensor by its L2 norm
    return x / (K.sqrt(K.mean(K.square(x))) + 1e-5)

#Define regularizations:
def blur_regularization(img, grads, size = (3, 3)):
    return cv2.blur(img, size)

def decay_regularization(img, grads, decay = 0.8):
    return decay * img

def clip_weak_pixel_regularization(img, grads, percentile = 1):
    clipped = img
    threshold = np.percentile(np.abs(img), percentile)
    clipped[np.where(np.abs(img) < threshold)] = 0
    return clipped

def gradient_ascent_iteration(loss_function, img, lr=0.9):
    loss_value, grads_value = loss_function([img])    
    gradient_ascent_step = img + grads_value * lr

    #Convert to row major format for using opencv routines
    grads_row_major = grads_value[0, :]
    img_row_major = gradient_ascent_step[0, :]

    #List of regularization functions to use
    regularizations = [blur_regularization, decay_regularization, clip_weak_pixel_regularization]

    #The reguarlization weights
    weights = np.float32([3, 3, 1])
    weights /= np.sum(weights)

    images = [reg_func(img_row_major, grads_row_major) for reg_func in regularizations]
    weighted_images = np.float32([w * image for w, image in zip(weights, images)])
    img = np.sum(weighted_images, axis = 0)

    #Convert image back to 1 x 3 x height x width
    img = np.float32([img])

    return img

def visualize_filter(input_img, filter_index, img_placeholder, layer, number_of_iterations = 20):
    loss = K.mean(layer[:, :, :, filter_index])
    grads = K.gradients(loss, img_placeholder)[0]
    grads = normalize(grads)
    # this function returns the loss and grads given the input picture
    iterate = K.function([img_placeholder], [loss, grads])

    img = input_img * 1

    # we run gradient ascent for 20 steps
    for i in range(number_of_iterations):
        img = gradient_ascent_iteration(iterate, img)

    # decode the resulting input image
    img = deprocess_image(img[0])
    #print("Done with filter", filter_index)
    return img

model = get_model(input_shape)
#model.summary()
model.load_weights(weights_path)
input_placeholder = model.input
layer = get_output_layer(model, layer_name)

if img_path is None:
    # we start from a gray image with some random noise
    init_img = np.random.random((1, img_width, img_height, 3)) * 20 + 128.
else:
    img = cv2.imread(img_path, 1)
    img = cv2.resize(img, (img_width, img_height))
    init_img = [img]

vizualizations = [None] * len(filter_indexes)
for i in tqdm(range(len(filter_indexes))):
    #for i,  in enumerate(filter_indexes):
    index = filter_indexes[i]
    vizualizations[i] = visualize_filter(init_img, index, input_placeholder,layer, iterations)
    #Save the visualizations see the progress made so far
    save_filters(vizualizations, img_width, img_height)
print('Done.')