juampamuc / deepvesselnet

Implementation of the DeepVesselNet deep learning network

Dependencies

1. keras
2. tensorflow / theano as backend for keras (tensoflow recommended)
3. sklearn
4. numpy

Installation

python setup.py install

Usage

from dvn import FCN, VNET, UNET  # import libraries

net = FCN()                                 # create the network object (You can replace FCN with VNET or UNET),
					    # there is a 'dim' parameter which takes the values 2, or 3 to build 2D or 3D versions of the networks (Default is 3)
					    # there is a 'cross_hair' parameter which builds a network with cross-hair filters when set to True (Default is False)

net.compile()                               # compile the network (supports keras compile parameters)
net.fit(x=X, y=Y, epochs=10, batch_size=10) # train the network (supports keras fit parameters)
preds = net.predict(x=X)                    # predict (supports keras predict parameters)
net.save(filename='model.dat')              # save network params
net = FCN.load(filename='model.dat')        # Load network params  (You can replace FCN with VNET or UNET as used above)

Datasets

For information on the data used for training etc. Have a look at the Datasets page in the wiki section