allenai / Holodeck

CVPR 2024: Language Guided Generation of 3D Embodied AI Environments.
https://yueyang1996.github.io/holodeck
Apache License 2.0
324 stars 30 forks source link
3d-environment ai2-thor generative-ai large-language-models text-to-3d unity


Language Guided Generation of 3D Embodied AI Environments


Paper | Project Page

Requirements

Holodeck is based on AI2-THOR, and we currently support macOS 10.9+ or Ubuntu 14.04+.

New Feature: To add ANY new assets to AI2-THOR, please check the objathor repo!

Note: To yield better layouts, use DFS as the solver. If you pull the repo before 12/28/2023, you must set the argument --use_milp to False to use DFS.

Installation

After cloning the repo, you can install the required dependencies using the following commands:

conda create --name holodeck python=3.10
conda activate holodeck
pip install -r requirements.txt
pip install --extra-index-url https://ai2thor-pypi.allenai.org ai2thor==0+8524eadda94df0ab2dbb2ef5a577e4d37c712897

Data

Download the data by running the following commands:

python -m objathor.dataset.download_holodeck_base_data --version 2023_09_23
python -m objathor.dataset.download_assets --version 2023_09_23
python -m objathor.dataset.download_annotations --version 2023_09_23
python -m objathor.dataset.download_features --version 2023_09_23

by default these will save to ~/.objathor-assets/..., you can change this director by specifying the --path argument. If you change the --path, you'll need to set the OBJAVERSE_ASSETS_DIR environment variable to the path where the assets are stored when you use Holodeck.

Usage

You can use the following command to generate a new environment.

python holodeck/main.py --query "a living room" --openai_api_key <OPENAI_API_KEY>

Our system uses gpt-4o-2024-05-13, so please ensure you have access to it.

Note: To yield better layouts, use DFS as the solver. If you pull the repo before 12/28/2023, you must set the argument --use_milp to False to use DFS.

Load the scene in Unity

  1. Install Unity and select the editor version 2020.3.25f1.
  2. Clone AI2-THOR repository and switch to the appropriate AI2-THOR commit.
    git clone https://github.com/allenai/ai2thor.git
    git checkout 07445be8e91ddeb5de2915c90935c4aef27a241d
  3. Reinstall some packages:
    pip uninstall Werkzeug
    pip uninstall Flask
    pip install Werkzeug==2.0.1
    pip install Flask==2.0.1
  4. Load ai2thor/unity as project in Unity and open ai2thor/unity/Assets/Scenes/Procedural/Procedural.unity.
  5. In the terminal, run this python script:
    python connect_to_unity --scene <SCENE_JSON_FILE_PATH>
  6. Press the play button (the triangle) in Unity to view the scene.

Citation

Please cite the following paper if you use this code in your work.

@InProceedings{Yang_2024_CVPR,
    author    = {Yang, Yue and Sun, Fan-Yun and Weihs, Luca and VanderBilt, Eli and Herrasti, Alvaro and Han, Winson and Wu, Jiajun and Haber, Nick and Krishna, Ranjay and Liu, Lingjie and Callison-Burch, Chris and Yatskar, Mark and Kembhavi, Aniruddha and Clark, Christopher},
    title     = {Holodeck: Language Guided Generation of 3D Embodied AI Environments},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2024},
    pages     = {16227-16237}
}