Keras MRI

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.

Install

pip install keras-mri

Usage

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/

Examples

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).

Development

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/*