This is a bunch of code to port Keras neural network model into pure C++. Neural network weights and architecture are stored in plain text file and input is presented as
vector<vector<vector<float> > > in case of image. The code is prepared to support simple Convolutional network (from MNIST example) but can be easily extended. There are implemented only ReLU and Softmax activations.
It is working with the Theano backend - support for Tensorflow will be added soon.
- Save your network weights and architecture.
- Dump network structure to plain text file with
- Use network with code from
keras_model.ccfiles - see example below.
- Run one iteration of simple CNN on MNIST data with
example/mnist_cnn_one_iteration.pyscript. It will produce files with architecture
example/my_nn_arch.jsonand weights in HDF5 format
- Dump network to plain text file
python dump_to_simple_cpp.py -a example/my_nn_arch.json -w example/my_nn_weights.h5 -o example/dumped.nnet.
- Compile example
g++ -std=c++11 keras_model.cc example_main.cc- see code in
- Run binary
./a.out- you shoul get the same output as in step one from Keras.
If you want to test dumping for your network, please use
test_run.sh script. Please provide there your network architecture and weights. The script do following job:
- Dump network into text file.
- Generate random sample.
- Compute predictions from keras and keras2cpp on generated sample.
- Compare predictions.