Linux installation steps

[glopesdev]

Installing Bonsai on Linux

1. Install the Mono framework for Linux.
2. Download the portable release of Bonsai 2.8 (.zip).
3. Extract the contents of the archive into a folder and run the following command mono Bonsai.exe.
4. Install Bonsai - Starter Pack from the package manager.
5. Go to the Updates tab in the package manager, select the Include prerelease checkbox and install all updates.

There is a common warning message related to libcanberra-gtk that can be removed by installing the relevant package:

sudo apt-get install libcanberra-gtk-module

Install OpenGL

Several Bonsai packages require OpenGL 4.0+ for interfacing with graphics. Below are some suggested instructions for a vanilla Ubuntu distribution. They may have to be tweaked for your particular system:

sudo apt install mesa-utils
glxinfo | grep "OpenGL version"
sudo apt install libsdl2-dev

The glxinfo command should return version 4.0 or greater for full compatibility with the shaders package.

Install OpenAL

The audio capture requires OpenAL as a dependency for interfacing with the microphone drivers.

sudo apt-get install libopenal1 libopenal-dev
sudo apt-get install libalut0 libalut-dev

Configure access to serial port

Occasionally the access to serial port on devices such as an Arduino might be blocked by your system. If this is the case, you can often use screen to configure access to the serial port:

screen /dev/ttyUSB0

Install OpenCV from pre-built binaries

Bonsai requires the C interface exposed by OpenCV 2.4 to drive its image processing primitives. OpenCV.NET provides pre-built binaries for a few ubuntu architectures on its latest release 3.4.2.

Add the earlier release repositories for libjasper

sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 3B4FE6ACC0B21F32
sudo apt update

Download and install pre-built binaries

The OpenCV project does not provide binary distributions for 2.4 for the latest versions of Ubuntu. These distributions are unofficial builds so you will have to download the .deb packages and install them locally.

wget https://github.com/horizongir/opencv.net/releases/download/v3.4.2/OpenCV-2.4.13.6-ubuntu-x86_64-libs.deb
wget https://github.com/horizongir/opencv.net/releases/download/v3.4.2/OpenCV-2.4.13.6-ubuntu-x86_64-dev.deb
sudo apt install OpenCV-2.4.13.6-ubuntu-x86_64-libs.deb
sudo apt install OpenCV-2.4.13.6-ubuntu-x86_64-dev.deb

Build OpenCV from source

You can also build OpenCV from source for your system. The below guide is adapted from the official build instructions.

Add the earlier release repositories for libjasper

sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 3B4FE6ACC0B21F32
sudo apt update

Install compiler dependencies

sudo apt-get install build-essential
sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev

Build and install OpenCV 2.4

Make sure to build the 2.4 branch, or download the tarball from GitHub.

For example to clone the 2.4 branch:

git clone -b 2.4 https://github.com/opencv/opencv.git

Example build configuration using cmake:

cd ~/opencv
mkdir release
cd release
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local ..

Build using make:

make
sudo make install

Configure environment

OpenCV installs libraries into usr/local/lib, so you may need to configure your environment using one of these options to share libraries at a system level.

[jbscherrer]

Is there a reason to expect Bonsai on linux to have better timing characteristics than on Windows? I'm trying to get feedback latency down as low as possible and wondering if switching to linux might be helpful.

[glopesdev]

With Mono I doubt you will see any improvements since Microsoft's implementation of generational GC was always superior. With the new .NET core however this picture might have changed, but we only have that update planned for next year's roadmap.

This Mono version was intended mostly as a preview and teaching aide but it does have the advantage of allowing to run Bonsai embedded on Raspberry Pis which might be useful for some simple embedded applications.

[roaldarbol]

Amazing work on Bonsai 2.8🥳 Can’t wait to run it on the Raspberry! Do you have a recommended way to install OpenCV for RPi? (I assume it’s slightly different from Ubuntu - or would you run Ubuntu instead of RPi OS?)

[ncguilbeault]

Updated - Building and installing OpenCV from source on Ubuntu 22.04

When trying to build OpenCV from source on Ubuntu 22.04, you may encounter some errors during the build process. I was able to successfully build and utilize OpenCV version 2.4.1 for Bonsai-Rx on Ubuntu 22.04 using the following updated steps.

Add old libjasper repository to package list:

sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 3B4FE6ACC0B21F32
sudo apt-key export C0B21F32 | sudo gpg --dearmour -o /etc/apt/trusted.gpg.d/libjasper.gpg
sudo apt update

Install compiler dependencies:

sudo apt-get install build-essential curl
sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev

OpenCV requires python 2, which no longer ships natively with Ubuntu. To install python 2 and numpy, you can run the following commands:

sudo apt-get install python2-dev

You then need to install the pip package manager for python 2, as well as the numpy library for python 2 using the following commands:

curl https://bootstrap.pypa.io/pip/2.7/get-pip.py --output get-pip.py
sudo python2 get-pip.py
sudo rm get-pip.py
python2 -m pip install numpy

You can install libdc with the following command:

sudo apt-get install libdc1394-dev

You can now clone the opencv repo:

git clone -b 2.4 https://github.com/opencv/opencv.git

And prepare to install using:

cd ~/opencv
mkdir release
cd release

This time, when running the make command, specify the paths to the python2 executable, include directory, and libraries. Specifying the c++ standard is also required:

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D PYTHON_EXECUTABLE=/usr/bin/python2 -D PYTHON_INCLUDE_DIR=/usr/include/python2.7 -D PYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython2.7.so -D CMAKE_CXX_STANDARD=11 ..

Finally, run to make the install script and install the OpenCV library:

make -j$(nproc)
sudo make install

For bonsai to access the OpenCV libraries, you can create an empty file to dynamically link the OpenCV library at the global system level:

sudo nano /etc/ld.so.conf.d/libopencv.conf

and add the following line to this file:

/usr/local/lib

Finally, run this command in the terminal to update:

sudo ldconfig

[chiyu1203]

Hi! @ncguilbeault I followed your post to install Bonsai on Ubuntu 22.04. The installation looked fine as I was able to go through the entire procedure. However, when I ran Bonsai and use the CameraCapture module, I bumped into this error when using OpenCV library. Have you seen error this before?
HIGHGUI ERROR: V4L/V4L2: VIDIOC_S_CROP
HIGHGUI ERROR: V4L2: Pixel format of incoming image is unsupported by OpenCV

[chiyu1203]

I uninstalled the old packages and recompiled everything with python3-based OpenCV based on @glopesdev original instruction (Build OpenCV from source) and solved the issue.

[ncguilbeault]

Hi @chiyu1203, glad you got it working. I have seen this error before as well. Installing the Video4Linux libraries solved this issue for me like so:

sudo apt update
sudo apt install v4l-utils libv4l-dev

[chiyu1203]

@ncguilbeault, thanks for putting bonsai-docker together! I have not tried that but it looks promising!

[caniko]

This is a great start for Linux support, but I would appreciate a dedicated page that is versioned and goes beyond Debian. I suggest we include guides for Fedora and Arch. I would love to have nix support, but understand that it is well beyond the scope of the project. Can we make this possible @glopesdev?

Opinion
I prefer mostly NixOS and Arch to a lesser extent over Ubuntu. I have created a request to support Bonsai on nix, please vote up if you are using nix.

[banchan86]

Since my laptop is a mac, and I have been working with documentation using docfx on mac, i decided for fun to try running bonsai with mono (V2.8.5, v2.9.0-preview1) on Mac and unfortunately it doesn't work because it looks for Nuget.Config in the wrong place:

NuGet.Configuration.NuGetConfigurationException: Failed to read NuGet.Config due to unauthorized access. Path: '/Users/XX/.config/NuGet/NuGet.Config'.

--→ I found it instead in /Users/XX/.nuget/NuGet/NuGet.Config

No real reason to fix it since I don't know who would use it outside of my very specific workcase haha and I am sure other stuff will break down the road, just wanted to put this somewhere in case anyone was wondering and as a followup to the last time this question was posted on the google groups discussion https://groups.google.com/g/bonsai-users/c/5E6KQPzvsRs. Haven't tried VMs.

[banchan86]

Got bonsai on mac running with @ncguilbeault bonsai dockers https://github.com/ncguilbeault/bonsai-docker/tree/dev!

[chiyu1203]

@banchan86 That's amazing! just curious, did you manage to run the bonsai-base image or bonsai-vision? If it was latter, did you install opencv2.4 via Homebrew or Macports?

[banchan86]

I only ran the bonsai-base image so don't have any experience with the bonsai-vision image unfortunately