Getting the ADAPT software stack
Below is a step-by-step explanation if you want to follow along or you can run this as a single script from here.
-
Install OS prerequisites:
sudo apt update sudo apt install dirmngr gnupg2 python-catkin-tools python-pip python-tk python-rosdep python-rosinstall \ python-vcstools ros-melodic-ros-core=1.4.1-0* ros-melodic-ros-base=1.4.1-0* ros-melodic-perception=1.4.1-0* \ ros-melodic-rqt-image-view libgl1-mesa-glx libqt5x11extras5 gdal-bin python-gdal python-ujson \ cuda-toolkit-10-2 cuda-core-10-2 libcudnn8-dev tensorrt tmux libhdf5-serial-dev hdf5-tools libhdf5-dev \ zlib1g-dev zip libjpeg8-dev liblapack-dev libblas-dev gfortran gpsd chrony pps-tools linuxptp \ nvidia-tensorrt python-libnvinfer-dev libopenblas-base libopenmpi-dev
-
Install python prerequisites:
# NOTE: install these dependencies as root so the systemd process can find them instead of a user account sudo pip install anaconda ipython ipykernel matplotlib spyder-kernels protobuf \ PyGeodesy shapely pyshp simplekml exifread future yacs scipy psutil pycuda sudo pip install Pillow --upgrade
-
Get this version of pytorch for tegra from NVIDIA:
cd ~ wget https://nvidia.box.com/shared/static/yhlmaie35hu8jv2xzvtxsh0rrpcu97yj.whl -O torch-1.4.0-cp27-cp27mu-linux_aarch64.whl sudo pip install future torch-1.4.0-cp27-cp27mu-linux_aarch64.whl
-
Setup torchvision
cd ~ git clone https://github.com/pytorch/vision cd vision git fetch origin v0.5.0 git reset --hard FETCH_HEAD sudo python setup.py install
-
Setup a specific version of torch2trt:
cd ~ git clone https://github.com/NVIDIA-AI-IOT/torch2trt.git cd torch2trt git fetch origin b0cc8e77a0fbd61e96b971a66bbc11326f77c6b5 git reset --hard FETCH_HEAD sudo python setup.py install --plugins
-
Run the rc_genicam_api install
cd ~ git clone https://github.com/roboception/rc_genicam_api.git cd rc_genicam_api mkdir build && cd build cmake -DCMAKE_INSTALL_PREFIX=/usr .. make && make package sudo dpkg -i rc-genicam-api*.deb
-
Setup the adapt ros workspace
cd ~ git clone https://gitlab.kitware.com/adapt/adapt_ros_ws.git cd adapt_ros_ws git submodule init git submodule update catkin build --force-cmake -DSETUPTOOLS_DEB_LAYOUT=OFF # get the model cd ~/adapt_ros_ws/models ./get_sim_model.sh
OPTIONAL items
Clean up some disk space:
sudo apt-get remove --purge libreoffice*
sudo apt-get autoclean
sudo apt-get autoremove
jtop for system stats:
pip install -U jetson-stats
# run jtop
jetsonUtilities:
git clone https://github.com/jetsonhacks/jetsonUtilities
cd jetsonUtilities
python jetsonInfo.py