Using gaze from Meta's aria glasses in for robot policy learning
For full official Aria documentation, see https://facebookresearch.github.io/projectaria_tools/docs/intro
- x64 Linux distributions of:
- Fedora 36, 37, 38
- Ubuntu focal (20.04 LTS) and jammy (22.04 LTS)
- Windows (MSVC 2019/2022)
- Mac Intel or Mac ARM-based (M1) with MacOS 11 (Big Sur) or newer
We provide instructions for both running our project code AND a full installation of the entire original codebase from Facebook (Meta) Research. For full installation from source, see "Installation from Official Source" below this installation guide.
- Clone the github repository onto your local machine
git clone https://github.com/UCLA-Robot-Intelligence-Lab/aria-gaze
- Create the anaconda environment:
conda env create -f environment.yml
- Activate the environemnt and install all libraries Make sure that you are in the right directory!
conda activate aria
pip install -r requirements.txt
To begin livestreaming on aria glasses, see these instructions
- Create virtual environment (using either venv or conda, both have been tested and work)
conda create -n aria python=3.9.20
conda activate aria
- Clone codebase (official aria codebase from Meta)
git clone https://github.com/facebookresearch/projectaria_tools.git -b 1.5.5
- the full codebase is only required for visualization, the actual livestreaming only requires common.py, test.py, and visualizer.py from the sdk files
- Install required python dependencies
python3 -m pip install --upgrade pip
python3 -m pip install projectaria-tools'[all]'
python3 -m pip install projectaria_client_sdk --no-cache-dir # Aria client SDK install (for livestreaming)
- Run aria-doctor to check for issues
aria-doctor
- Pair glasses to computer
aria auth pair
After running this command in terminal, go to the Aria app on your phone and approve the pairing
- Extract code for livestreaming
python -m aria.extract_sdk_samples --output ~
cd ~/projectaria_client_sdk_samples
python3 -m pip install -r requirements.txt