Quickstart#

Get visualkeras running in just 5 minutes!

Installation#

Install visualkeras using pip:

pip install visualkeras

Your First Visualization#

Let’s create a simple neural network and visualize it using the unified show() API:

import tensorflow as tf
from tensorflow import keras
import visualkeras

# Create a simple Sequential model
model = keras.Sequential([
    keras.layers.Input(shape=(28, 28, 1)),
    keras.layers.Conv2D(32, (3, 3), activation='relu'),
    keras.layers.MaxPooling2D((2, 2)),
    keras.layers.Conv2D(64, (3, 3), activation='relu'),
    keras.layers.MaxPooling2D((2, 2)),
    keras.layers.Conv2D(64, (3, 3), activation='relu'),
    keras.layers.Flatten(),
    keras.layers.Dense(64, activation='relu'),
    keras.layers.Dense(10)
])

# Display the model architecture using show() with mode parameter
image = visualkeras.show(model, mode='layered')
image.show()

What Just Happened?#

visualkeras generated a layered view of your model (using mode='layered'), showing:

  • Each layer as a colored block

  • Tensor shapes flowing through the network

  • How data is transformed at each step

The show() function is the unified API that supports four visualization modes.

This layered view is great for understanding Convolutional Neural Networks (CNNs).

Using Presets#

visualkeras includes curated presets for different use cases:

# Compact layout (minimal spacing)
compact = visualkeras.show(model, mode='layered', preset='compact')

# Presentation style (large, detailed, suitable for papers/slides)
presentation = visualkeras.show(model, mode='layered', preset='presentation')

# Default balanced style
default = visualkeras.show(model, mode='layered', preset='default')

Trying Different Modes#

The show() API makes it easy to experiment with different visualization styles:

# These all work with the same function - just change the mode

layered_viz = visualkeras.show(model, mode='layered')        # Layer stacks
graph_viz = visualkeras.show(model, mode='graph')            # Computational graph
functional_viz = visualkeras.show(model, mode='functional')  # Layered + graph hybrid
lenet_viz = visualkeras.show(model, mode='lenet')             # Classic feature map style

Try each mode to see which best represents your model!

Next Steps#

Ready to explore more? Check out:

For more control over styling, see Tutorials.

Common Customizations#

Changing Colors and Styles#

Use the unified show() API with color customization:

# Define colors with fill and outline for each layer type
image = visualkeras.show(
    model,
    mode='layered',
    color_map={
        keras.layers.Conv2D: {'fill': '#3498db', 'outline': '#2980b9'},
        keras.layers.MaxPooling2D: {'fill': '#2ecc71', 'outline': '#27ae60'},
        keras.layers.Dense: {'fill': '#e74c3c', 'outline': '#c0392b'}
    }
)
image.show()

Using Different Visualization Modes#

visaalkeras supports four visualization modes. Each mode is useful for different model types:

# Layered view: Great for CNNs, shows stacked feature maps
layered = visualkeras.show(model, mode='layered')

# Graph view: Shows computational structure for any model type
graph = visualkeras.show(model, mode='graph')

# Functional view: Graph-aware layering (bridges both approaches)
functional = visualkeras.show(model, mode='functional')

# LeNet view: Classic feature map stack visualization
lenet = visualkeras.show(model, mode='lenet')

# Use presets for publication-ready output
publication = visualkeras.show(model, mode='layered', preset='presentation')

Saving to File#

image = visualkeras.layered_view(model)
image.save('model_architecture.png')

Troubleshooting#

The visualization is too crowded

  • Try using graph_view() instead of layered_view()

  • Reduce the scale factor: visualkeras.layered_view(model, scale_xy=20)

Shapes aren’t displaying correctly

  • Make sure your model has an Input layer defined

  • Call model.build((batch_size, *input_shape)) if needed

Getting an error?

Need Help?#