https://github.com/phtrempe/conviz

Conviz is a convolutional neural network layer visualization library developed in Python and used with Keras.
https://github.com/phtrempe/conviz

convolutional-layers convolutional-networks convolutional-neural-networks keras python python-3 python3 tensorflow theano visualization

Last synced: 5 days ago
JSON representation

Conviz is a convolutional neural network layer visualization library developed in Python and used with Keras.

• Host: GitHub
• URL: https://github.com/phtrempe/conviz
• Owner: PhTrempe
• License: mit
• Created: 2017-03-22T19:51:45.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-04-04T21:05:44.000Z (over 7 years ago)
• Last Synced: 2024-09-29T09:19:51.431Z (about 1 month ago)
• Topics: convolutional-layers, convolutional-networks, convolutional-neural-networks, keras, python, python-3, python3, tensorflow, theano, visualization
• Language: Python
• Homepage: https://github.com/PhTrempe/conviz
• Size: 1.25 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


    






| Better Code Hub compliance |

|:--------------------------:|

| [![Better Code Hub compliance](https://bettercodehub.com/edge/badge/PhTrempe/conviz)](https://bettercodehub.com/) |

## Description

Conviz is a convolutional neural network layer visualization library developed 

in Python and used with Keras.

## How to Install

```

pip install conviz

```

## A Small Example to Get You Started

```python

from conviz.models import cifar10

from conviz.utils.image_util import ImageUtil

from conviz.visualizer import Visualizer

# Loads a model trained on the CIFAR10 dataset

model = cifar10.load()

# Creates and binds a visualizer to the model

visualizer = Visualizer(model)

# Gets the layer of the model to visualize

layer = model.get_layer("conv1")

# Generates the visualization for the selected layer as a 4 by 4 grid of filters

img = visualizer.visualize(layer, (4, 4), ga_rate=0.1, num_steps=1000)

# Saves the generated visualization image to a file

ImageUtil.save_image(img, "./conv1.png")

```