Menoh is DNN inference library with C API.

Menoh is released under MIT License.

DISCLAIMER: Menoh is still experimental. Use it at your own risk. In particular not all operators in ONNX are supported, so please check whether the operators used in your model are supported. We have checked that VGG16 and ResNet50 models converted by onnx-chainer work fine.

Document

This codebase contains C API and C++ API.

For Windows users, prebuild libraries are available (see release) and Nuget package is available.

See also

• Chainer model to ONNX : onnx-chainer
• C# wrapper : menoh-sharp
• Haskell wrapper : menoh-haskell
• [Unofficial] Go wrapper by kou-m san : gomenoh
• [Unofficial] Rust wrapper by Y-Nak san : menoh-rs
• [Unofficial] Rust wrapper by Hakuyume san : menoh-rs

• DNN Inference with CPU
• ONNX support
• Easy to use.

• MKL-DNN Library (0.14 or later)
• Protocol Buffers (3.5.1 is checked)

Execute below commands in root directory.

python retrieve_data.py
mkdir build && cd build
cmake ..
make

Execute below command in build directory created at Build section.

make install

Execute below command in root directory.

./example/vgg16_example_in_cpp

Result is below

vgg16 example
-22.3708 -34.4082 -10.218 24.2962 -0.252342 -8.004 -27.0804 -23.0728 -7.05607 16.1343
top 5 categories are
8 0.96132 n01514859 hen
7 0.0369939 n01514668 cock
86 0.00122795 n01807496 partridge
82 0.000225824 n01797886 ruffed grouse, partridge, Bonasa umbellus
97 3.83677e-05 n01847000 drake

Please give --help option for details

./example/vgg16_example_in_cpp --help

Setup chainer

Then, execute below commands in root directory.

python gen_test_data.py
cd build
cmake -DENABLE_TEST=ON ..
make
./test/menoh_test.out

• Elu
• LeakyRelu
• Relu
• Softmax
• Tanh

• Concat

• Conv
• ConvTranspose
• FC

• Abs
• Add
• Sqrt

• BatchNormalization

• AveragePool
• GlobalAveragePool
• GlobalMaxPool
• MaxPool

Menoh is released under MIT License. Please see the LICENSE file for details.

Note: retrieve_data.py downloads data/VGG16.onnx. data/VGG16.onnx is generated by onnx-chainer from pre-trained model which is uploaded at http://www.robots.ox.ac.uk/%7Evgg/software/very_deep/caffe/VGG_ILSVRC_16_layers.caffemodel

That pre-trained model is released under Creative Commons Attribution License.