Skip to content

Latest commit

 

History

History
160 lines (130 loc) · 8.56 KB

README.md

File metadata and controls

160 lines (130 loc) · 8.56 KB
Raw

[NeurIPS 2023] LayoutGPT: Compositional Visual Planning and Generation with Large Language Models

Weixi Feng1*, Wanrong Zhu1*, Tsu-Jui Fu1, Varun Jampani3, Arjun Akula3, Xuehai He2, Sugato Basu3, Xin Eric Wang2, William Yang Wang1
1UC Santa Barbara, 2UC Santa Cruz, 3Google
*Equal Contribution

Project Page | arxiv

Teaser figure

Example 1 Example 2 Example 3

Updates

2024.03.18 Resolved 3D data preparation issues and added a script for rendering with blender.

2023.10.28 Now support Llama-2; camera ready version updated

2023.10.10 We released our preprocessed 3D-FRONT and 3D-FUTURE data (see below). Simplified the installation and preparation process.

2023.09.22 LayoutGPT is accepted to NeurIPS 2023!

Installation & Dependencies

LayoutGPT and the downstream generation requires different libraries. You can install everything all at once

conda create -n layoutgpt python=3.8 -y
pip install -r requirements.txt

and additionally

# for GLIGEN
wget https://huggingface.co/gligen/gligen-generation-text-box/resolve/main/diffusion_pytorch_model.bin -O gligen/gligen_checkpoints/checkpoint_generation_text.pth

# for image evaluation using GLIP
cd eval_models/GLIP
python setup.py build develop --user
wget https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window12_384_22k.pth -O MODEL/swin_large_patch4_window12_384_22k.pth
wget https://huggingface.co/GLIPModel/GLIP/resolve/main/glip_large_model.pth?download=true -O MODEL/glip_large_model.pth

# for scene synthesis
cd ATISS
python setup.py build_ext --inplace
pip install -e .

You may also refer to the official repo of GLIGEN, GLIP and ATISS for detailed guidance.

Data Preparation

Our image layout benchmark NSR-1K and the 3D scene data split is provided under ./dataset.

2D image layouts

NSR-1K contains ground truth image layouts for each prompt extracted from the MSCOCO dataset. The extracted clip image features are provided under ./dataset/NSR-1K/. The json files contain ground truth layouts, captions and other metadata.

3D scene layouts

Download 3D-FUTURE and our preprocessed data to ./ATISS/. Then unzip these files.

cd ATISS
unzip 3D-FUTURE-model.zip -d 3D-FUTURE
unzip data_output.zip

2D Image Layout Generation

We provide the script to generate layouts for NSR-1K benchmark. First set up your openai authentication in the script. Then run

python run_layoutgpt_2d.py --icl_type k-similar --K 8 --setting counting --llm_type gpt4 --n_iter 5

The generated layout will be saved to ./llm_output/counting by default. To generate images based on the layouts, run

cd gligen
python gligen_layout_counting.py --file ../llm_output/counting/gpt4.counting.k-similar.k_8.px_64.json --batch_size 5

Note that the script will save a clean image and an image with bounding boxes for each prompt into two separate folders. In our experiment in the preprint, we generate 5 different layouts for each prompt to reduce variance.

Layout & Image Evaluation

To evaluate the raw layouts, run

# for numerical prompts
python eval_counting_layout.py --file ../llm_output/counting/gpt4.counting.k-similar.k_8.px_64.json

To evaluate the generated images using GLIP, run

cd eval_models/GLIP
python eval_counting.py --dir path_to_generated_clean_images

3D Indoor Scene Synthesis

First set up your openai authentication in the script, then run the script to generate scenes

python run_layoutgpt_3d.py --dataset_dir ./ATISS/data_output --icl_type k-similar --K 8 --room bedroom --llm_type gpt4 --unit px --normalize --regular_floor_plan

To evaluate the out-of-bound rate (OOB) and KL divergence (KL-div.) of the generated layouts, run

python eval_scene_layout.py --dataset_dir ./ATISS/data_output --file ./llm_output/3D/gpt4.bedroom.k-similar.k_8.px_regular.json --room bedroom

Blender Visualization

Run the following command to generte necessary files and have a low-quality visualization of the scene:

cd ATISS/scripts
python render_from_files.py ../config/bedrooms_eval_config.yaml ../visualization ../data_output_future/threed_future_model_bedroom.pkl ../demo/floor_plan_texture_images ../../llm_output/3D/gpt4.bedroom.k-similar.k_8.px_regular.json --without_screen --up_vector 0,1,0 --camera_position 2,2,2 --split test_regular --export_scene

With --export_scene, object and material files for each scene will be saved to a folder in ./ATISS/visualization/. Make sure you download Blender and can execute from command line (Linux&Windows: extract .tar.xz/.zip, Mac: install .dmg and then make an alias).

# example
blender -b -P render_with_blender.py -- --input_dir ../visualization/test_Bedroom-803 --output_dir ../visualization/test_Bedroom-803.png --camera_position 0 0 5

Citation

Please consider citing our work if you find it relevant or helpful:

@article{feng2024layoutgpt,
  title={Layoutgpt: Compositional visual planning and generation with large language models},
  author={Feng, Weixi and Zhu, Wanrong and Fu, Tsu-jui and Jampani, Varun and Akula, Arjun and He, Xuehai and Basu, Sugato and Wang, Xin Eric and Wang, William Yang},
  journal={Advances in Neural Information Processing Systems},
  volume={36},
  year={2024}
}

Disclaimer

We thank the authors of GLIGEN, GLIP and ATISS for making their code available. It is important to note that the code present here is not the official or original code of the respective individual or organization who initially created it. Part of the code may be subject to retraction upon official requests. Any use of downstream generation code should be governed by the official terms and conditions set by the original authors or organizations. It is your responsibility to comply with these terms and conditions and ensure that your usage adheres to the appropriate guidelines.