Installing NNStreamer on macOS

DISCLAIMER: This manual is work-in-progress and needs to be updated. Specifically, only PyTorch support is covered for now.

Prerequisites

The following dependencies are needed for installing and building:

• Apple's Clang >= 10.0.0
• Homebrew for installing dependencies

Installing via Homebrew

The simplest way to install NNStreamer on macOS is through our third-party Homebrew repository:

$ brew tap nnstreamer/neural-network
$ brew install nnstreamer

Building from source

Note that most frameworks will be disabled during the configuration, if you manage to make NNStreamer work on this platform with your preferred framework, please provide a pull request with an update of this doc!

Install the necessary dependencies:

$ brew install meson ninja pkg-config cmake libffi glib \
    gstreamer gst-plugins-base gst-plugins-good numpy

If you need to run PyTorch models:

$ brew install libtorch

Clone NNStreamer's repository:

$ git clone https://github.com/nnstreamer/nnstreamer.git
$ cd nnstreamer

Optionally checkout a recent version more stable than the main branch, for instance:

$ git checkout v2.1.1

Configure the build with meson:

$ meson build \
    --prefix=/usr/local \
    -Dwerror=false -Denable-test=false

The target files will be installed in /usr/local/lib/gstreamer-1.0, /usr/local/lib/nnstreamer, /usr/local/include/nnstreamer and /usr/local/etc/.

Finally build and install using ninja:

$ ninja -C build install