Skip to content
How to run Keras model on RK3399Pro
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
freeze_graph.ipynb add, edit readme May 2, 2019

Run Keras/Tensorflow model on RK3399Pro

Clone or download this repo

git clone

Download pre-compiled Python wheel files from my aarch64_python_packages repo and rknn_toolkit wheels from their official GitHub.

Step1: Freeze Keras model and convert to RKNN model (On Linux development machine)

Require Python 3.5+.

Install required libraries for your development machine

pip3 install -r requirements.txt

The install rknn toolkit with the following command.

pip3 install rknn_toolkit-0.9.9-cp36-cp36m-linux_x86_64.whl

To freeze a Keras InceptionV3 ImageNet model to a single .pb file. The frozen graph will accept inputs with shape (N, 299, 299, 3).

To convert the .pb file to .rknn file, run


Step2: Make prediction (On RK3399Pro board)

Setup for the first time.

sudo dnf update -y
sudo dnf install -y cmake gcc gcc-c++ protobuf-devel protobuf-compiler lapack-devel
sudo dnf install -y python3-devel python3-opencv python3-numpy-f2py python3-h5py python3-lmdb
sudo dnf install -y python3-grpcio

sudo pip3 install scipy-1.2.0-cp36-cp36m-linux_aarch64.whl
sudo pip3 install onnx-1.4.1-cp36-cp36m-linux_aarch64.whl
sudo pip3 install tensorflow-1.10.1-cp36-cp36m-linux_aarch64.whl
sudo pip3 install rknn_toolkit-0.9.9-cp36-cp36m-linux_aarch64.whl

To run inference benchmark on RK3399Pro board, in its terminal run,

You can’t perform that action at this time.