Installing Ubuntu 20.04 on raspberry pi for Tensorflow

Publish date: Dec 2, 2020

Tags:

raspberry

Why Ubuntu - 32 bit/ 64 bit ?

Out of plenty of operating systems available for raspberry pi, we need stable one’s intended to run ml projects (tf-lite, opencv). Raspberry Pi group’s debian based raspbian os provide good support and stable, but it is 32 bit. There is a 64 bit version out but in beta stage. We need a 64 bit OS for increased performance. Ubuntu provides OS in 64 bit version for rpi. It’s good to go.

Headless installation

If you don’t have separate monitor, (or) keyboard, (or) mouse for the regular installation you can choose this process.

Download the Ubuntu Server Image from the official site: link
Download balena etcher and flash the image to the SD card.
After flashing is done, open boot folder from the sd card and edit network-config file to connect wifi.

Uncomment the lines below and change according your network.

wifis:
  wlan0:
  dhcp4: true
  optional: true
  access-points:
    <wifi network name>:
      password: "<wifi password>"

Save the file and remove sd card and place it in raspberry pi. After first boot wait for 20 sec and restart the machine.

Check for rpi’s ip address using wifi portal or fing app in mobile phone and ssh into the machine using the ip address.

Desktop flavour install

We installed the server, next step is to install desktop environment. There are flavours in ubuntu for DE. As we have less computational resources, choose a lightweight flavour like xubuntu/lubuntu.

I used this awesome script to install desktop environment.

  
  git clone https://github.com/wimpysworld/desktopify.git
  
  cd desktopify
  
  sudo ./desktopify --de xubuntu

Finally install xrdp server to access the gui remotely. (xrdp »> vnc)

  
  sudo apt-get install xrdp

you can connect remotely using remmina by ip-address

overclocking

Add these below lines at end of /boot/config.txt

  
  sudo nano /boot/config.txt

over_voltage = 2
arm_freq = 1750

then reboot : - )

To check clock speed use this command:

  
  `lscpu | grep MHz.`

memory swapping

  
  sudo apt-get install zram-config
  
  sudo nano /usr/bin/init-zram-swapping

Add a multiplier (* 5) at the end of the mem…. line. This gives you about 4.5 Gbyte of memory swap ( like this 1024 * 5)

reboot : - )

tensorflow install

Build tensorflow using the pre-built binaries from this repo: https://github.com/PINTO0309/Tensorflow-bin
Build tensorflow-lite using the pre-built binaries from this repo: https://github.com/PINTO0309/TensorflowLite-bin

opencv install

Option 1

directly installing from ppa repo:

  
  sudo apt-get update
  
  sudo apt-get upgrade
  
  sudo apt install libopencv-dev python3-opencv

you can verify the installation by following command:

  
  python3 -c "import cv2; print(cv2.__version__)"

Option 2 - directly from source

Before moving forward remember to set swap space(> 3gb), it is essential for build process.

Install these following dependencies

  
  sudo apt-get update
  
  sudo apt-get upgrade
  
  sudo apt-get install build-essential cmake gcc g++ git unzip pkg-config
  
  sudo apt-get install libjpeg-dev libpng-dev libtiff-dev
  
  sudo apt-get install libavcodec-dev libavformat-dev libswscale-dev
  
  sudo apt-get install libgtk2.0-dev libcanberra-gtk*
  
  sudo apt-get install libxvidcore-dev libx264-dev
  
  sudo apt-get install python3-dev python3-numpy python3-pip
  
  sudo apt-get install libtbb2 libtbb-dev libdc1394-22-dev
  
  sudo apt-get install libv4l-dev v4l-utils
  
  sudo apt-get install libopenblas-dev libatlas-base-dev libblas-dev
  
  sudo apt-get install liblapack-dev gfortran libhdf5-dev
  
  sudo apt-get install libprotobuf-dev libgoogle-glog-dev libgflags-dev
  
  sudo apt-get install protobuf-compiler

Next, download and unzip the packages

  
  cd ~
  
  wget -O opencv.zip https://github.com/opencv/opencv/archive/4.4.0.zip
  
  wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/4.4.0.zip
  
  unzip opencv.zip
  
  unzip opencv_contrib.zip
  
  #rename the folders
  
  mv opencv-4.4.0 opencv
  
  mv opencv_contrib-4.4.0 opencv_contrib
  
  cd opencv
  
  mkdir build
  
  cd build

Build make

$ cmake -D CMAKE_BUILD_TYPE=RELEASE \
	  -D CMAKE_INSTALL_PREFIX=/usr/local \
	  -D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib/modules \
	  -D ENABLE_NEON=ON \
	  -D BUILD_TIFF=ON \
      -D WITH_FFMPEG=ON \
	  -D WITH_GSTREAMER=ON \
	  -D WITH_TBB=ON \
      -D BUILD_TBB=ON \
      -D BUILD_TESTS=OFF \
      -D WITH_EIGEN=OFF \
      -D WITH_V4L=ON \
      -D WITH_LIBV4L=ON \
      -D WITH_VTK=OFF \
      -D OPENCV_ENABLE_NONFREE=ON \
      -D INSTALL_C_EXAMPLES=OFF \
      -D INSTALL_PYTHON_EXAMPLES=OFF \
      -D BUILD_NEW_PYTHON_SUPPORT=ON \
      -D BUILD_opencv_python3=TRUE \
      -D OPENCV_GENERATE_PKGCONFIG=ON \
      -D BUILD_EXAMPLES=OFF ..

build with these commands

  
  make -j4
  
  sudo make install
  
  sudo ldconfig
  
  sudo apt-get update

Verify installation

  
  pkg-config --modversion opencv4
  
  python3 -c "import cv2; print(cv2.__version__)"