Installation¶

AEStream is usable both as a command-line binary or Python tool.

Source	Installation
pip	`pip install aestream`
nix	`nix run github:aestream/aestream` (CLI) `nix develop github:aestream/aestream` (Python environment)
docker	See Installation documentation

Contributions to support AEStream on additional platforms are always welcome.

Installing via pip¶

Run pip install aestream

Installing via pip is the most convenient method for installing AEStream and gives you both the command-line interface and the Python API. Pip is the Python package manager and is accessible on most computers.

Installing with CUDA support¶

If you want to use AEStream with CUDA support, you likely need to build the package yourself. This is because the CUDA version must match the one installed on your system. To do so, simply run pip install aestream --no-binary aestream to avoid using the binary cache. It will take a few minutes to compile, but if you have an NVIDIA GPU, you’ll see dramatic performance improvements. Note that you can provide a -v flag to enable verbose output, which will show you if the CUDA drivers were detected (look for CUDA found).

Event camera drivers¶

AEStream can read from Inivation or Prophesee event cameras, given that the drivers are installed. Please make sure the drivers are installed before installing aestream (step 1 below).

Follow the instructions at https://gitlab.com/inivation/dv/libcaer/ to install the Inivation drivers and/or https://github.com/prophesee-ai/openeb/ to install the Prophesee drivers
Install AEStream with pip: pip install aestream --no-binary aestream -v
- The --no-binary flag forces pip to recompile aestream for your system, which will detect the event camera drivers, if present.
- The -v flag enables verbose output, which will show you if the drivers were detected
- If you already have an installation, you need to re-install
Ensure that the drivers were detected by inspecting the installation log
- If successful, you should see messages like -- Inivation dependencies (libcaer) found at /usr/local/lib/cmake/libcaer
Test your installation by running the usb_video.py example. If a window with streaming events pops up, you are successful!

Once installed, the cameras can be used in both CLI or Python.

We are working to automate the driver installation as a part of AEStream. Help is welcome and appreciated!

Installing via Docker¶

A Dockerfile is available in the root directory and can be built with the Dockerfile available in the aestream root directory as follows:

docker build -t aestream .

Installing via source¶

Source installations can either be done via pip, which wraps the C++ code, or CMake, where you have to manually configure the project.

Python source install and development¶

git clone https://github.com/aestream/aestream
cd aestream
pip install .

Note that this will setup the install in an isolated virtual environment, which is slow and can cause problems if you would like to work with the source code. To avoid this, we recomment using --no-build-isolation in the pip install, but that requires manually installing the necessary packages for the build system, shown below.

# Pull the source code
git clone https://github.com/aestream/aestream
cd aestream
# Install build dependencies
pip install scikit-build-core setuptools_scm pathspec
# Build aestream, but without build isolation
# Any future compilation will *only* compile the files you changed
pip install --no-build-isolation .

CMake source install¶

You can also install the C++ code directly using CMake.

git clone https://github.com/aestream/aestream
cd aestream
mkdir build
cd build
cmake -GNinja .. 
ninja install

Note that this will only install the CLI version of AEStream. To have access to the Python features, you must enable the USE_PYTHON config in CMAKE by running cmake -GNinja -DUSE_PYTHON=ON .. instead.

If your default C++ compiler doesn’t support C++ 20, you will have to install an up-to-date compiler and provide the environmental variable CXX. For instance like this: CXX=/path/to/g++-10 cmake -GNinja ..

Requirements¶

AEStream relies on modern compiler features and require at least GCC 10 and CMake 3.20.

If you are on older machines that default to GCC versions below 10, you can enforce a specific version by setting the environmental variable CXX. Here is an example for pip:

CXX=/path/to/g++-10 pip install aestream

If you have problems with CMake, try installing the latest version from pip: pip install cmake.

Common issues¶

Here, we list some common issues that users encounter during installation. If you run into problems that is not listed here, please open an issue.

CMake Error: Could not find a package configuration file provided by “nanobind” with any of the following names: …

This happens when the nanobind dependency could not be found. Run pip install nanobind to install it and retry.