This User Quick Guide will help you get started to setup MXNet-HRT on RK3399 quickly.

2. Preparation

2.1 General dependencies installation

sudo apt-get update -y
sudo apt-get upgrade -y
sudo apt-get install build-essential git libatlas-base-dev  libblas-dev libopencv-dev -y 
sudo apt-get install python-pip python-dev -y
sudo apt-get install -y python-numpy python-scipy
sudo pip install --upgrade pip
sudo apt-get install scons –y
sudo apt-get install git –y

2.2 Download source code

cd ~

Download "AID-tools" (AID-tools : v1.0):

wget ftp://ftp.openailab.net/tools/package/AID-tools.tar.gz

Download "MXNet-HRT" :

git clone --recursive https://github.com/OAID/MXNet-HRT.git

3. Build MXNet-HRT

3.1 install AID-tools :

sudo tar -xvf AID-tools.tar.gz -C /usr/local
sudo /usr/local/AID/gen-pkg-config-pc.sh /usr/local/AID

3.2 Build MXNet :

cd ~/MXNet-HRT
make
sudo make install
sudo /usr/local/AID/gen-pkg-config-pc.sh /usr/local/AID

3.3 Build Classification Sample

cd example/image-classification/predict-cpp
export CXX=aarch64-linux-gnu-g++
export USE_ACL=1
make

4. Run Caffenet Classification

4.1 Download MXNet Model

cd ~/MXNet-HRT
cd model
   You can download MXNet pretrained model and synset text from  http://data.mxnet.io/mxnet/models/imagenet/
   or download model from ftp://ftp.openailab.net/tools/Caffe-HRT_test_model/models.tar.gz

4.2 Run MXNet Classification

cd ~/MXNet-HRT
example/image-classification/predict-cpp/image-classification-predict-forCaffeMode cpu model/caffenet/caffenet-symbol.json model/caffenet/caffenet-0000.params model/Inception/mean_224.nd model/synset.txt model/pictures/cat.jpg
The output message:
Best Result: [ tabby, tabby cat] id = 281, accuracy = 0.27792722