Visualizing Vision Models in Keras

Summarizing a Model

In Keras, the summary() method is used to generate the summary of the provided model. It returns a string which contains a detailed summary of the model.

The summary() method provides a textual overview about:

• The number of layers and their order in the model.
   
• The shape of each layer and the number of weights present in each layer.
   
• The total number of parameters present in the model.
   

Example

Let us now create a basic neural network using Keras to summarize it using the summary() method.

Code

#Importing libraries
from keras.models import Sequential
from keras.layers import Dense

#Creating a simple sequential model
model = Sequential()
model.add(Dense(2, input_dim=1, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.add(Dense(2, activation='softmax'))

#Calling summary method
print(model.summary())

 

Output

Explanation

In the above example, we imported the sequential class and dense layer from Keras. Then we created a simple sequential model in which we added three dense layers with various activation functions. Finally, we called the summary() method on the model, which gives us the overall summary of the model structure.

Visualizing a Model

Visualizing neural networks is very vital as it can help us to understand the structure and links inside the neural network. Simple models can be visualized easily, but when complex models are used, then visualization becomes very difficult. In this scenario, we can use various methods available in the Keras library.

The below methods can be used to visualize the model in Keras:

• plot_model() method
• Using Visualkeras
   

Let us now discuss both methods one by one.

Using plot_model() method

The plot_model() method is available in tensorflow.keras.utils module. It allows you to plot and visualize any Keras module.

Syntax

plot_model (model, to_file, show_shapes, show_layer_names, dpi)

 

Description of syntax:

• model: It is the Keras model that you want to visualize.
   
• to_file: It is the file name of the plot image.
   
• show_shapes: (Optional) It specifies whether to show shape information. By default, it is set to false.
   
• show_layer_names: (Optional) It specifies whether to show the names of each layer. By default, it is set to true.
   
• dpi: (Optional) It is the resolution (dots per inch) of the saved plot.
   

Example

Now let us use the previous sequential model and visualize it using the plot_model() method.

Code

#Importing libraries
from keras.models import Sequential
from keras.layers import Dense

#Importing plot_model
from keras.utils.vis_utils import plot_model

#Creating a simple sequential model
model = Sequential()
model.add(Dense(2, input_dim=1, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.add(Dense(2, activation='softmax'))

#Calling plot_model method
plot_model(model, to_file='plot_model.png', show_shapes=True, show_layer_names=True, dpi=100)

 

Output

 

Explanation

In the above example, we imported the plot_model() from keras.utils. After that, we plotted our model with the required parameters for the plot_model() method and got our required result.

Using Visualkeras

The visualkeras is an open-source Python library used to visualize neural network models. It supports a layered and graph-type architecture of neural networks, and thus it is ideal for CNN models. Let’s see the command to install visualkeras.

Command

pip install visualkeras

Example

Now, let’s build a convolution neural network model and visualize it using visualkeras library.

Code

#Importing libraries
from tensorflow import keras
from tensorflow.keras import layers, models

#Creating a sequential model and adding layers
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='softmax', input_shape=(36, 36, 3)))
model.add(layers.MaxPooling2D((2, 2)))

model.add(layers.Conv2D(32, (3, 3), activation='softmax', input_shape=(36, 36, 3)))
model.add(layers.MaxPooling2D((2, 2)))

model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))

model.add(layers.Dense(64, activation='softmax'))
model.add(layers.Flatten())

#Calling summary method
model.summary()

 

The summary of the above model is given as follows:

Now to visualize the above model, replace the below lines to your code with summary() method.

import visualkeras
visualkeras.layered_view(model) 

 

Here, the visulakeras show the convolution layer in yellow, the pooling layer in pink, the dense layer in green and flatten layer in blue colour. 

We can label these layers using the ImageFont library. Add the below code to import and add labels in your visualization.

from PIL import ImageFont
font = ImageFont.truetype("arial.ttf", 16)
visualkeras.layered_view(model, legend=True, font=font) 

 

Here we are visualizing our model in 3D space. But visulakeras also allows us to visualize a model in 2D space. 

In order to add 2D visualization, make the value of the draw_volume argument false.

visualkeras.layered_view(model, legend=True, font=font, draw_volume=False) 

 

Similarly, there are other features available in the visualkeras library for you to explore. 

Frequently Asked Questions

What is Keras?

Keras is an open-source deep learning library in Python. It provides high-level neural network API to build deep learning models.

Why do we need to visualize models in Keras?

Visualizing models helps us understand the model architecture better, allowing us to analyze and optimize our model easily.

How to visualize models in Keras?

The plot_model() method in Keras is used to create a visual sketch of models giving explicit info about their layers and links.

What is the summary() method in Python?

The summary() method in Python is a method which is used to get a detailed overview of a model or data frame.

Conclusion

This article discusses the topic of visualizing vision models in Keras. We discussed different methods through which we can visualize a deep learning model in Python. We hope this blog has helped you enhance your knowledge of visualizing vision models in Keras. If you want to learn more, then check out our articles.

• Application of Keras
• Transfer Learning with Keras
• Convolutional Neural Network in Keras
   

Refer to our Guided Path to upskill yourself in DSA, Competitive Programming, JavaScript, System Design, and many more! If you want to test your coding ability, you may check out the mock test series and participate in the contests hosted on Coding Ninjas!

But suppose you have just started your learning process and are looking for questions from tech giants like Amazon, Microsoft, Uber, etc. In that case, you must look at the problems, interview experiences, and interview bundles for placement preparations.

However, you may consider our paid courses to give your career an edge over others!

Happy Learning!