MAC.md

Install pyslam under macOS

• Install pyslam under macOS
  • Install steps
  • Usage
  • Notes about macOS
    • Install homebrew
  • Known issues
    • First run of the scripts
    • Issues found with OpenCV and pyenv
    • Issues found with boost serialization

The following procedure has been tested under Sequoia 15.1.1 and Xcode 16.1.

Install steps

Please, follow these install steps:

1. First you need to install homebrew and Xcode CLT (check the section below)
2. Download this repo:

   git clone --recursive https://github.com/luigifreda/pyslam.git 
   cd pyslam 

3. Change your default shell type to bash:

   chsh -s /bin/bash 

   (if you want to set zsh back then run: chsh -s /bin/zsh)
4. Launch the install script

   ./install_all.sh  # unified install procedure

NOTE: The above install procedure (step 4) installs a virtual python environment pyslam in your system. That virtual environment can be easily activated by using the command:

. pyenv-activate.sh 

(do not forget the dot without '/' ! ) You can find further details about python virtual environments here.

Usage

To test the basic visual odometry run

. pyenv-activate.sh   # Activate pyslam environment. This is just needed once in a new terminal.
./main_vo.py

To test the full SLAM system run

. pyenv-activate.sh   # Activate pyslam environment. This is just needed once in a new terminal.
./main_slam.py

Notes about macOS

Install homebrew

From https://brew.sh/, run the following command:

/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"

You may also need to install gcc by using XCode Command Line Tools. Run

xcode-select --install

Known issues

First run of the scripts

On my machine, I've noticed that, the first time one of main scripts is run, it takes a while before it actually starts.

Issues found with OpenCV and pyenv

When you launch one of the scripts above, you get a warning:

objc[6169]: Class CaptureDelegate is implemented in both /Users/luigi/.python/venvs/pyslam/lib/python3.7/site-packages/cv2/cv2.cpython-37m-darwin.so (0x11923d590) and /usr/local/opt/opencv/lib/libopencv_videoio.4.3.dylib (0x13021d0c8). One of the two will be used. Which one is undefined.

This is an open issue which needs to be solved. However, this is not a show stopper.

In a few words, this is an "interference" between the OpenCV libs of the installed virtual python environment and the OpenCV libs installed by homebrew.

Issues found with boost serialization

On my mac, boost deserialization is very slow. On the other hand, under linux, it is very fast.