Counting 3,567 Big Data & Machine Learning Frameworks, Toolsets, and Examples...
Suggestion? Feedback? Tweet @stkim1

Last Commit
Feb. 22, 2019
Jul. 15, 2016


Build Status (License MIT 1.0)


Use Keras models in C++ with ease

Table of contents


Would you like to build/train a model using Keras/Python? And would you like run the prediction (forward pass) on your model in C++ without linking your application against TensorFlow? Then frugally-deep is exactly for you.


  • is a small header-only library written in modern and pure C++.
  • is very easy to integrate and use.
  • depends only on FunctionalPlus, Eigen and json - also header-only libraries.
  • supports inference (model.predict) not only for sequential models but also for computational graphs with a more complex topology, created with the functional API.
  • re-implements a (small) subset of TensorFlow, i.e. the operations needed to support prediction.
  • results in a much smaller binary size than linking against TensorFlow.
  • works out of-the-box also when compiled into a 32-bit executable. (Of course 64 bit is fine too.)
  • utterly ignores even the most powerful GPU in your system and uses only one CPU core. ;-)
  • but is quite fast on one CPU core compared to TensorFlow.

Supported layer types

Layer types typically used in image recognition/generation are supported, making many popular model architectures possible (see Performance section).

  • Add, Concatenate, Subtract, Multiply, Average, Maximum
  • AveragePooling1D/2D, GlobalAveragePooling1D/2D
  • Bidirectional, Embedding, GRU, LSTM, TimeDistributed
  • Conv1D/2D, SeparableConv2D, DepthwiseConv2D
  • Cropping1D/2D, ZeroPadding1D/2D
  • BatchNormalization, Dense, Flatten
  • Dropout, AlphaDropout, GaussianDropout
  • SpatialDropout1D, SpatialDropout2D, SpatialDropout3D
  • MaxPooling1D/2D, GlobalMaxPooling1D/2D
  • ELU, LeakyReLU, ReLU, SeLU, PReLU
  • Sigmoid, Softmax, Softplus, Tanh
  • UpSampling1D/2D
  • Reshape, Permute

Also supported

  • multiple inputs and outputs
  • nested models
  • residual connections
  • shared layers
  • variable input shapes
  • arbitrary complex model architectures / computational graphs

Currently not supported are the following:

ActivityRegularization, AveragePooling3D, Conv2DTranspose, Conv3D, ConvLSTM2D, CuDNNGRU, CuDNNLSTM, Cropping3D, Dot, GaussianNoise, GRUCell, Lambda, LocallyConnected1D, LocallyConnected2D, LSTMCell, Masking, MaxPooling3D, RepeatVector, RNN, SimpleRNN, SimpleRNNCell, StackedRNNCells, ThresholdedReLU, Upsampling3D, any custom layers, temporal models


  1. Use Keras/Python to build (model.compile(...)), train ( and test (model.evaluate(...)) your model as usual. Then save it to a single HDF5 file using'....h5', include_optimizer=False). The image_data_format in your model must be channels_last, which is the default when using the TensorFlow backend. Models created with a different image_data_format and other backends are not supported.

  2. Now convert it to the frugally-deep file format with keras_export/

  3. Finally load it in C++ (fdeep::load_model(...)) and use model.predict(...) to invoke a forward pass with your data.

The following minimal example shows the full workflow:

import numpy as np
from keras.layers import Input, Dense
from keras.models import Model

inputs = Input(shape=(4,))
x = Dense(5, activation='relu')(inputs)
predictions = Dense(3, activation='softmax')(x)
model = Model(inputs=inputs, outputs=predictions)
model.compile(loss='categorical_crossentropy', optimizer='nadam')
    np.asarray([[1,2,3,4], [2,3,4,5]]),
    np.asarray([[1,0,0], [0,0,1]]), epochs=10)'keras_model.h5', include_optimizer=False)
python3 keras_export/ keras_model.h5 fdeep_model.json
// main.cpp
#include <fdeep/fdeep.hpp>
int main()
    const auto model = fdeep::load_model("fdeep_model.json");
    const auto result = model.predict(
        {fdeep::tensor5(fdeep::shape5(1, 1, 1, 1, 4), {1, 2, 3, 4})});
    std::cout << fdeep::show_tensor5s(result) << std::endl;

When using a test case (input and corresponding output values) is generated automatically and saved along with your model. fdeep::load_model runs this test to make sure the results of a forward pass in frugally-deep are the same as in Keras.

Some integration examples


Below you can find the average durations of multiple consecutive forward passes for some popular models ran on a single core of an Intel Core i5-6600 CPU @ 3.30GHz. frugally-deep was compiled (GCC ver. 5.4.0) with g++ -O3 -mavx (same as TensorFlow binaries). The processes were started with CUDA_VISIBLE_DEVICES='' taskset --cpu-list 1 ... to disable the GPU and to only allow usage of one CPU.

Model Keras + TF frugally-deep
DenseNet121 0.96 s 0.32 s
DenseNet169 1.17 s 0.35 s
DenseNet201 1.50 s 0.46 s
InceptionV3 0.71 s 0.38 s
MobileNet 0.34 s 0.16 s
MobileNetV2 0.40 s 0.16 s
NASNetLarge 4.22 s 4.73 s
NASNetMobile 0.34 s 0.38 s
ResNet50 0.73 s 0.27 s
VGG16 0.66 s 0.78 s
VGG19 0.82 s 0.97 s
Xception 1.50 s 1.20 s

Keras version: 2.2.2

TensorFlow version: 1.10.1

Requirements and Installation

A C++14-compatible compiler is needed. Compilers from these versions on are fine: GCC 4.9, Clang 3.7 (libc++ 3.7) and Visual C++ 2015.

Guides for different ways to install frugally-deep can be found in


frugally-deep uses channels_last (height, width, depth/channels) as its image_data_format internally, as does TensorFlow. Everything is handled as a float-32 tensor with rank 5.

In case you would like to use double instead of float for all calculations, simply do this:

#define FDEEP_FLOAT_TYPE double
#include <fdeep/fdeep.hpp>

A frugally-deep model is thread-safe, i.e. you can call model.predict on the same model instance from different threads simultaneously. This way you may utilize up to as many CPU cores as you have predictions to make. With model::predict_multi there is a convenience function available to handle the parallelism for you.


The API of this library still might change in the future. If you have any suggestions, find errors or want to give general feedback/criticism, I'd love to hear from you. Of course, contributions are also very welcome.


Distributed under the MIT License. (See accompanying file LICENSE or at

Latest Releases
 Feb. 22 2019
 Jan. 19 2019
 Jan. 13 2019
 Jan. 5 2019
 Dec. 2 2018