# create conda environment with python 3.6
conda create --name tf python=3.6
conda activate tf

# Install conda Jupyter notebooks supports.
conda install ipython nb_conda_kernels

# Install tensorflow, cuda and cudnn toolkit. We use cuda10 as example.
conda install -c anaconda tensorflow-gpu cudatoolkit=10.0

# Install dependences
pip install cython
pip install Keras==2.2.4 tqdm Pillow scikit-image opencv-python h5py imgaug pycocotools requests

# Clone this repo
git clone https://github.com/Jiang-Murray/Keras_Model_Wrapper.git
cd Keras_Model_Wrapper

# Activate environment for tensorflow applications
conda activate tf

# Run Pascal_Voc trained DeepLabv3 Xception backbone on demo video
python deeplab_video.py xception

# Run Pascal_Voc trained DeepLabv3 Mobilenetv2 backbone on demo video
python deeplab_video.py mobilenetv2

# Run COCO trained Mask_RCNN demo on video
python mask_rcnn_video.py

# Run classification demo on imagenet
python imagenet_demo.py

# Run Pascal_Voc trained DeepLabv3 Xception backbone on ROS topic /camera/left/image_raw/compressed and publish a compressed color map image to /deeplab/semantic/compressed
python ros/deeplab_node.py xception /camera/left/image_raw/compressed /deeplab/semantic/compressed

# Run Pascal_Voc trained DeepLabv3 Mobilenetv2 backbone on ROS topic /camera/left/image_raw/compressed and publish a uncompressed color map image to /deeplab/semantic/
python ros/deeplab_node.py mobilenetv2 /camera/left/image_raw/compressed /deeplab/semantic/

# Run COCO trained Mask_RCNN on ROS topic /camera/left/image_raw/compressed
python ros/maskrcnn_node.py /camera/left/image_raw/compressed