Simple guideline for installing all the main computer vision libraries in the SBC Jetson Nano
At least a 32 GB SD-Card (I'm using a 64 GB)
USB fan to reduce CPU Tempeature to avoid CPU throttle.
sudo fallocate -l 4G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
sudo swapon --show
sudo cp /etc/fstab /etc/fstab.bak
echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab
sudo apt install -y git
sudo apt install -y cmake
sudo apt install -y libatlas-base-dev
sudo apt install -y gfortran
sudo apt install -y python3-dev
sudo apt install -y python3-pip
sudo apt install -y libhdf5-serial-dev
sudo apt install -y hdf5-tools
sudo apt install mlocate
pip3 install numpy (NOT WORKING)
pip3 install -U pip setuptools
pip3 install matplotlib (NOT WORKING)
sudo apt-key adv --keyserver keys.gnupg.net --recv-key F6E65AC044F831AC80A06380C8B3A55A6F3EFCDE || sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-key F6E65AC044F831AC80A06380C8B3A55A6F3EFCDE
sudo add-apt-repository "deb http://realsense-hw-public.s3.amazonaws.com/Debian/apt-repo xenial main" -u\
sudo apt-get update
sudo apt-get install librealsense2-dkms
sudo apt-get install librealsense2-utils
sudo apt-get install librealsense2-dev
sudo apt-get install librealsense2-dbg
Test it by typping:
realsense-viewer
sudo pip3 install python-dev-tools
sudo apt-get install python3.6-devel
sudo apt update && sudo apt upgrade
sudo apt install htop
wget https://raw.githubusercontent.com/AastaNV/JEP/master/script/install_opencv4.1.1_Jetson.sh
mkdir opencv
git clone https://github.com/AastaNV/JEP.git
cd JEP/script/
./install_opencv4.1.1_Jetson.sh ~/opencv [Takes some time (2h-4h), depends of power source and cooler]
Add to the .bashrc file the following:
export PYTHONPATH=$PYTHONPATH:/<where-you-place-jep-folder>/JEP/script/opencv-4.1.1/release/python_loader/
import opencv should work in python
git clone https://github.com/intel-isl/Open3D.git
util/scripts/install-deps-ubuntu.sh
mkdir build
cd build
cmake ..
sudo make -j4
sudo make install-pip-package (wheel)
import open3d should work in python
sudo apt-get dist-upgrade
sudo apt-get install python3 python3-dev
git clone https://github.com/IntelRealSense/librealsense.git
cd librealsense
mkdir build
cd build
cmake ../ -DBUILD_PYTHON_BINDINGS:bool=true
make -j4
sudo make install
export PYTHONPATH=$PYTHONPATH:/usr/local/lib (or place it in the ./bashrc file)
import pyrealsense2 should work in python
sudo pip3 install keyboard
import keyboard should work in python
wget https://nvidia.box.com/shared/static/2ls48wc6h0kp1e58fjk21zast96lpt70.whl -O torch-1.0.0a0+bb15580-cp36-cp36m-linux_aarch64.whl
sudo pip3 install numpy torch-1.0.0a0+bb15580-cp36-cp36m-linux_aarch64.whl
cd /usr/local/lib/python3.6/site-packages or /<virtual-env-dir>/lib/python3.6/site-packages
git clone https://github.com/pytorch/vision
cd vision
sudo python3 setup.py install
or
sudo pip3 install torchvision
add to the bashrc file the following:
export CUDADIR=/usr/local/cuda-10.0
export PATH=$PATH:/usr/local/cuda-10.0/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-10.0/lib64
sudo apt install terminator
sudo apt-get install ninja-build
python3 setup.py install
Add the following lines to the .bashrcfile
My ninja directory is \usr\bin\ninja and cuda directory is \usr\local\cuda-10.0 but it can depend.
export PATH=<your-ninja-directory>:${PATH}
export CUDA_HOME=<your-cuda-directory>
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${LD_LIBRARY_PATH}
git clone https://github.com/fxia22/pointnet.pytorch.git
cd pointnet.pytorch
pip3 install -e .
cd scripts
bash build.sh
bash download.sh
sudo apt-get install python3-matplotlib
sudo pip3 install scipy
Dependencies:
pip3 install pydensecrf
pip3 install protobuf
pip3 install tensorboardX
pip3 install imageio
Download data for desired dataset.
Download repository: git clone https://github.com/PingoLH/FCHarDNet
Create setup file: gedit config.yaml
Extract the zip and modify the path in the config.yaml file.
To train the model:
python3 train.py --config [config.yaml]
sudo apt install i3
sudo reboot
Select option in the login screen - i3
Shortcuts:
windowns + Shift + q -> close selected application
windowns + Enter -> open new terminal
windowns + d -> open search menu
windowns + (number of workspace 0-9) -> change workspace
windowns + shift + (nº of workspace) -> move/send to workspace
windowns + f -> fullscreen current aplication
windowns + r -> resize selected windown -> Use arrows to resize and "Esc" to leave the resize mode
Configuration file to create new shortcuts:
home/<your-user>/.config/i3/config
sudo gedit /etc/gdm3/custom.conf
uncomment the two lines below # Enabling automatic login
Add the following to the ~/.bashrc file:
echo Running python script
python3 your_python_script.py
What is left is to open the terminal after boot to run the script.
cd ~/.config/i3
add the following to the config file:
for_window [class="Terminal"] move container to workspace 1
exec --no-startup-id i3-sensible-terminal
The installation of DIGITS must be made by source because the POP!OS! system is incompatible with the nvidia docker.
Follow these instructions: https://github.com/NVIDIA/DIGITS/blob/digits-6.0/docs/BuildDigits.md
The command ./digits-devserver may not work, because of the flask_socketio package.
Remove the flask_socketio package from the requeriments.txt inside digits folder.
Install the last version by doing: pip install flask_socketio .
Them you might have another error, to solve upgrade the werkzeug library, pip install --upgrade werkzeug==0.16.1
To start the server simply use the command: ./digits-devserver \
In case of error in detection models, just do: sudo pip install --user --upgrade protobuf==3.1.0.post1
First use command df -h to see the available devices.
Plug the SDCard in your computer and use the command df -h again.
Now you should see a new line and that's your SDCard, normally it looks like /dev/mmcblk0p1
Use sudo dd if=/dev/mmcblk0 of=~/SDCardBackupJetpack43.img
This previous command will save the current image (jetpack4.3) of your SD-Card in /SDCardBackupJetpack43.img
It will take some time (around 20~30 minutes).
For formating the SDCard use gparted and remove all the partitions and format in fat32
What's left to do is to download the image from: https://developer.nvidia.com/embedded/jetpack-archive\ And follow the Jetson Nano instructions: https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit#write