tspannhw / onnxinferencejava

Convert a Chainer model to ONNX format, import to MXNet and perform inference using MXNet's Java API

ONNX model inference using MXNet's Java API

Recommendations:

• Use Python 3.5+ as onnx_chainer requires Python 3.5+
• Chainer is not guaranteed to work on MacOS and Windows (Reference)
• Use conda to create a virtual environment for Python

Installation:

1. Clone this repo with

git clone https://github.com/vandanavk/ONNXinferenceJava.git
cd ONNXinferenceJava/

2. Create and activate a conda environment

pip install conda
conda create -n mxnet_blog python=3.6

conda activate mxnet_blog

Tip: If you are unable to activate the conda environment, try this

3. Install Python dependencies

pip install mxnet==1.4.0
pip install chainer==5.3.0
pip install chainercv==0.12.0
pip install onnx==1.3.0
pip install onnx_chainer==1.3.3

4. For Java, install the pre-requisites Reference

For macOS

brew update
brew tap caskroom/versions
brew cask install java8
brew install maven

For Ubuntu

sudo apt-get install openjdk-8-jdk maven

Code execution:

1. Execute the Python file to convert the Chainer model to an MXNet model

python chainer_onnx_mxnet.py

Check model/ folder for

chainer_vgg16.onnx, vgg16-symbol.json, vgg16-0000.params, synset.txt

Check data/ folder for Penguin.jpg

2. Execute the Java file to perform inference

mvn clean dependency:copy-dependencies install
java -Xmx8G -cp target/ONNXJava-1.0-SNAPSHOT.jar:target/dependency/* mxnet.ONNXMXNetJava --model-path-prefix model/vgg16 --input-image data/Penguin.jpg

Alternatively, open ONNXInferenceJava in an IDE, build the project and run ONNXMXNetJava.java

Deactivate conda environment

conda deactivate mxnet_blog