Training Toolbox for TensorFlow provides a convenient environment to train Deep Learning models and convert them using OpenVINO™ Toolkit for optimized inference.
- Ubuntu 16.04 / 18.04
- Python 3.4-3.6
- libturbojpeg
- For Python pre-requisites refer to
requirements.txt - (Optional) TensorFlow GPU pre-requisites.
- (Optional) OpenVINO™ R3 for exporting of the trained models
- Clone repository in the working directory
cd /<path_to_working_dir>
git clone https://github.com/opencv/training_toolbox_tensorflow.git
cd training_toolbox_tensorflow
git submodule update --init --recursive
- Install prerequisites
sudo apt-get install libturbojpeg python3-tk python3-pip virtualenv 2to3
bash tools/init_venv.sh
- Start to work
cd /<path_to_working_dir>/training_toolbox_tensorflow
. venv/bin/activate
In virtual environment run tests:
cd training_toolbox
nosetests
or if you are going to use the OpenVino toolkit:
cd training_toolbox
export OPEN_VINO_DIR=<PATH_TO_OPENVINO>
nosetests
Note: if you have installed the OpenVino toolkit after creating a virtual environment then you have to recreate one to install required packages for the Model Optimizer into one.
Do not forget to update several environment variables are required to compile and run OpenVINO™ toolkit applications, for details see: https://software.intel.com/en-us/articles/OpenVINO-Install-Linux.
After installation, you are ready to train your own models, evaluate them, use them for predictions.