yu4u/convnet-drawer

Python script for illustrating Convolutional Neural Networks (CNN) using Keras-like model definitions

convnet-drawer's Issues

AttributeError: 'LSTM' object has no attribute 'filters'

AttributeError: 'LSTM' object has no attribute 'filters'.How to solve this problem?

Unable to save model as pptx file

Hi there,

Another error to declare:

  File "C:\Users\ngenne\Desktop\PRIVE-Deep-Learning\NIH\chest-xray\convnet_drawer\pptx_util.py", line 63, in save_model_to_pptx
    for feature_map in model.feature_maps + model.layers:

AttributeError: 'Sequential' object has no attribute 'feature_maps'

Code is below:

from convnet_drawer.convnet_drawer import Model, Conv2D, MaxPooling2D, Flatten, Dense
from convnet_drawer.keras_util import convert_drawer_model
from convnet_drawer.matplotlib_util import save_model_to_file
from convnet_drawer.pptx_util import save_model_to_pptx
from keras.models import Sequential
from convnet_drawer.keras_models import AlexNet

classifier = Sequential()

classifier = Model(input_shape=(1024, 1024, 1))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(MaxPooling2D(pool_size=(2, 2)))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(MaxPooling2D(pool_size=(2, 2)))
classifier.add(Flatten())
classifier.add(Dense(4096))
classifier.add(Dense(4096))
classifier.add(Dense(1000))

classifier = AlexNet.get_model()
classifier_seq = convert_drawer_model(classifier)

classifier.save_model_to_pptx(classifier, "classifier.pptx")

Could you help me please?

Matplotlib_util get an error

Hi there,

I'm getting an error everytime I want to generate the graph into the pdf file:

  File "C:\Users\ngenne\Desktop\PRIVE-Deep-Learning\NIH\chest-xray\convnet_drawer\matplotlib_util.py", line 10, in save_model_to_file
    plt.xlim(model.x, model.x + model.width)

AttributeError: 'Sequential' object has no attribute 'x'

Below my code:

from convnet_drawer.convnet_drawer import Model, Conv2D, MaxPooling2D, Flatten, Dense
from convnet_drawer.keras_util import convert_drawer_model
from convnet_drawer.matplotlib_util import save_model_to_file
from keras.models import Sequential
from convnet_drawer.keras_models import AlexNet

classifier = Sequential()



classifier = Model(input_shape=(1024, 1024, 1))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(MaxPooling2D(pool_size=(2, 2)))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(Conv2D(512, (256,256), strides=(4, 4), padding="same"))
classifier.add(MaxPooling2D(pool_size=(2, 2)))
classifier.add(Flatten())
classifier.add(Dense(4096))
classifier.add(Dense(4096))
classifier.add(Dense(1000))

classifier = AlexNet.get_model()
classifier_seq = convert_drawer_model(classifier)

Do you have a trick for me?

TypeError: super() takes at least 1 argument (0 given)

Running the example (AlexNet.py) I get the error "TypeError: super() takes at least 1 argument (0 given" from the line "model = Model(input_shape=(224, 224, 3))." Any ideas?

Missing Keras Layers

The Dropout-Layer doesnt seem to be represented.
Error comes from the keras_util
NameError: name 'Dropout' is not defined

Unable to create a multi-head model

So I was wondering if its possible to visualize a multihead model [1], as I can only define one Input.
Is there anything i missed out?

[1] https://3qeqpr26caki16dnhd19sv6by6v-wpengine.netdna-ssl.com/wp-content/uploads/2018/07/Plot-of-the-Multi-Headed-1D-Convolutional-Neural-Network.png

ImportError: No module named 'convnet_drawer'

I've installed python-pptx, but this error occurred while I was running：
from convnet_drawer import Model, Conv2D, Deconv2D
ImportError: No module named 'convnet_drawer'

AttributeError: 'str' object has no attribute 'get'

I receive an error about 'str' object has no attribute 'get':

from keras.models import Sequential
from keras.layers import Dense, Conv2D, Flatten, Dropout, MaxPooling2D

model = Sequential()

model.add(Conv2D(16, kernel_size=3, strides=3, activation='relu',
data_format="channels_last", input_shape=(129,1173,1)))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(32, kernel_size=3, strides=3, activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, kernel_size=3, strides=3, activation='relu'))
model.add(Flatten())
model.add(Dense(units=256, activation='relu'))
model.add(Dense(units=128, activation='relu'))
model.add(Dense(units=64, activation='relu'))
model.add(Dense(8, activation='relu'))
model.add(Dense(1))
model.compile(loss='mae', optimizer='adam')

model_draw = convert_drawer_model(model)
model_draw.save_fig("11.svg")

---

AttributeError Traceback (most recent call last)
in
----> 1 model_draw = convert_drawer_model(model)
2 model_draw.save_fig("11.svg")

~/Notebooks/convnet-drawer/keras_util.py in convert_drawer_model(model)
33 figure = Model(input_shape=_input_shape[1:])
34 for config in model.get_config():
---> 35 class_name = config.get("class_name", False)
36 class_config = config.get("config", False)
37 if class_name and class_config:

AttributeError: 'str' object has no attribute 'get'

Run error

Hi, I follow every step of the instruction, however, does not work, anyone can help me, thank you!

Traceback (most recent call last):
File "D:/lyceum/plot_CNN_architecture/CNN_archi.py", line 23, in
save_model_to_pptx(model, "1.pptx")
File "D:\lyceum\plot_CNN_architecture\pptx_util.py", line 63, in save_model_to_pptx
presentation.add_line(obj.x1, obj.y1, obj.x2, obj.y2, obj.color, obj.width, obj.dasharray)
File "D:\lyceum\plot_CNN_architecture\pptx_util.py", line 29, in add_line
connector.ln = connector.get_or_add_ln()
AttributeError: can't set attribute

I really love your work, however

I really love your work, as I said earlier but why don't you add the other layers, (Batch Normalization, Activation, Dropout, Zero Padding 2D, Max Pooling 2D), or at least package your code?

I am ready and willing to contribute with you, if you want..

convert_drawer_model fail when GlobalAveragePooling2D layer is included

Hi there,
I have come across some problem. convert_drawer_model from keras_util fail when GlobalAveragePooling2D layer is included. But it is said here that GlobalAveragePooling2D is supported.
Thanks in advance.

update

fix this problem. see pr #12

my code

from keras_util import convert_drawer_model
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense,Dropout,Activation,Flatten
from keras.layers import Conv2D,MaxPool2D,GlobalAveragePooling2D,AveragePooling2D
from keras import optimizers
from keras.callbacks import ReduceLROnPlateau,EarlyStopping
from keras.layers.normalization import BatchNormalization
from keras.regularizers import l2

batch_size = 128
epochs = 30
num_classes = 10
weight_decay = 1e-6
nets = 15

model = Sequential()

model.add(Conv2D(192,(5,5),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal',input_shape=(28,28,1)))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(160,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(96,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(MaxPool2D(pool_size=(3,3),strides=(2,2),padding='same'))

#model.add(Dropout(0.2))

model.add(Conv2D(192,(5,5),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(192,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(192,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(MaxPool2D(pool_size=(3,3),strides=(2,2),padding='same'))

#model.add(Dropout(0.2))

model.add(Conv2D(192,(3,3),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(192,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))
model.add(Conv2D(10,(1,1),padding='same',kernel_regularizer=l2(weight_decay),kernel_initializer='he_normal'))
#model.add(BatchNormalization())
#model.add(Activation('relu'))  

model.add(GlobalAveragePooling2D())
#model.add(Activation('softmax'))

#adam = optimizers.rmsprop()
#model.compile(loss='categorical_crossentropy',optimizer=adam,metrics=['accuracy'])

Error

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-19-677489d1360a> in <module>
----> 1 convert_drawer_model(model).save_fig('tmp.svg')

~/python_lib/convnet_drawer/keras_util.py in convert_drawer_model(model)
     36         class_config = config.get("config", False)
     37         if class_name and class_config:
---> 38             class_obj = is_class_object(class_name)
     39             if class_name == "Conv2D":
     40                 conv_2d = get_conv2d_obj(class_obj, class_config)

~/python_lib/convnet_drawer/keras_util.py in is_class_object(class_name)
     26 
     27 def is_class_object(class_name):
---> 28     return eval(class_name)
     29 
     30 

~/python_lib/convnet_drawer/keras_util.py in <module>

NameError: name 'GlobalAveragePooling2D' is not defined

ImportError: No module named 'convnet_drawer'

Getting import error after installing via pip... Can anyone please help?

Export PPTX issue/question

Hi there,

Thanks for making this great project! I am interested in using the project to export CNN diagrams in powerpoint, however I am running into an issue I was wondering if you maybe know how to troubleshoot.

I downloaded AlexNet.pptx from your repo and the graph looks like its not being rendered on the powerpoint correctly, here is a screenshot of what I am seeing in powerpoint:

It seems that the render location is going beyond the white region area, and going off the slides. Is this an issue on my powerpoint file, or do I have to offset the location in order to have the diagram rendered correctly.

When I export it to jpg, the rendering issue exists.

Please advise if you need more information, and I look forward to your response!

Is there any plan to support Deconvolution2D() ?