Keras MRI is a neural network visualization tool for Keras. This is still in Beta and some networks (e.g. RNNs and 3D CNNs) and types of layers are not yet supported.

pip install keras-mri

import kmri

from keras.models import Sequential
from keras.layers import Dense

model = Sequential([
    Dense(8, input_shape=(3,))
])

input_batch = [[1,2,3], [7,8,9]]
kmri.visualize_model(model, input_batch)

Once this is running navigate to http://127.0.0.1:5000/

The examples folder contains models that demonstrate keras-mri usage. To run one of the examples run the following from one of the example folders

pip install -r requirements.txt
python main.py

chrome-dino

Simple reinforcement learning model that learned to play the dinosaur game in chrome. Demonstrates visualizing a convolutional neural network with multiple inputs. This is based on chrome-dino-ai.

img-classifier

Demonstrates using keras-mri to visualize standard image classifieres (VGG, MobileNet, and ResNet).

Install package locally

pip install -e .

Deploy

# increase version number in setup.py
# Move unrelease changes to next version in changelog.md
# commit, tag with vx.x.x, and push
# Delete build, dist, and egg-info
python setup.py sdist bdist_wheel
python -m twine upload dist/*