The DreamCraft3D extension of threestudio. The original implementation can be found at https://github.com/deepseek-ai/DreamCraft3D. We thank them for their contribution to the 3D generation community. To use it, please install threestudio first and then install this extension in threestudio custom directory.
cd custom
git clone https://github.com/DSaurus/threestudio-dreamcraft3D.git
# If you want to use your custom image, please install background remover
pip install backgroundremover
You also need to make sure that you have an access token from huggingface to use DeepFloyd-IF.
huggingface-cli login
To download stable-zero123, please go to the load/zero123 directory and run download.sh.
# It will take about 3 hours and ~22G GPU memory to run all stages
prompt="a delicious hamburger"
image_path="load/images/hamburger_rgba.png"
# --------- Stage 1 (NeRF & NeuS) --------- #
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-nerf.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path"
ckpt=outputs/dreamcraft3d-coarse-nerf/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-neus.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.weights="$ckpt"
# --------- Stage 2 (Geometry Refinement) --------- #
ckpt=outputs/dreamcraft3d-coarse-neus/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-geometry.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
# --------- Stage 3 (Texture Refinement) --------- #
ckpt=outputs/dreamcraft3d-geometry/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-texture.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
# It will take about 40min to run all stages
prompt="a delicious hamburger"
image_path="load/images/hamburger_rgba.png"
# --------- Stage 1 (NeRF & NeuS) --------- #
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-nerf-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path"
ckpt=outputs/dreamcraft3d-coarse-nerf/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-neus-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.weights="$ckpt"
# --------- Stage 2 (Geometry Refinement) --------- #
ckpt=outputs/dreamcraft3d-coarse-neus/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-geometry-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
# --------- Stage 3 (Texture Refinement) --------- #
ckpt=outputs/dreamcraft3d-geometry/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-texture-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
prompt="a delicious hamburger"
image_path="load/images/hamburger_rgba.png"
ckpt=path/to/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-YOURSTAGE.yaml --export system.prompt_processor.prompt="$prompt" data.image_path="$image_path" resume="$ckpt" system.exporter.context_type=cuda
To run with your custom image, first you need preprocess images to remove background and get depth/normal maps. You can also get image caption with the following script.
# preprocess images
cd custom/threestudio-dreamcraft3D
python image_preprocess.py "examples/hamburger.png" --size 512 --border_ratio 0.0
# if you need image caption
# python image_preprocess.py "examples/hamburger.png" --size 512 --border_ratio 0.0 --need_caption
# if you remove backgounrd using other tools like ClipDrop, you need put the processed image to "examples/hamburger_rgba.png" and add option use_existing_background
# python image_preprocess.py "examples/hamburger.png" --size 512 --border_ratio 0.0 --use_existing_background
cd ../..
You will get the results, including rgba, depth and normal images.
Then you can run dreamcraft3D using the following script.
prompt="a delicious hamburger"
image_path="custom/threestudio-dreamcraft3D/examples/hamburger_rgba.png"
# --------- Stage 1 (NeRF & NeuS) --------- #
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-nerf.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path"
ckpt=outputs/dreamcraft3d-coarse-nerf/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-neus.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.weights="$ckpt"
# --------- Stage 2 (Geometry Refinement) --------- #
ckpt=outputs/dreamcraft3d-coarse-neus/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-geometry.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
# --------- Stage 3 (Texture Refinement) --------- #
ckpt=outputs/dreamcraft3d-geometry/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-texture.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
Or you can run with fast config:
prompt="a delicious hamburger"
image_path="custom/threestudio-dreamcraft3D/examples/hamburger_rgba.png"
# --------- Stage 1 (NeRF & NeuS) --------- #
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-nerf-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path"
ckpt=outputs/dreamcraft3d-coarse-nerf/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-coarse-neus-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.weights="$ckpt"
# --------- Stage 2 (Geometry Refinement) --------- #
ckpt=outputs/dreamcraft3d-coarse-neus/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-geometry-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
# --------- Stage 3 (Texture Refinement) --------- #
ckpt=outputs/dreamcraft3d-geometry/$prompt@LAST/ckpts/last.ckpt
python launch.py --config custom/threestudio-dreamcraft3D/configs/dreamcraft3d-texture-fast.yaml --train system.prompt_processor.prompt="$prompt" data.image_path="$image_path" system.geometry_convert_from="$ckpt"
If you find DreamCraft3D helpful, please consider citing:
@article{sun2023dreamcraft3d,
title={Dreamcraft3d: Hierarchical 3d generation with bootstrapped diffusion prior},
author={Sun, Jingxiang and Zhang, Bo and Shao, Ruizhi and Wang, Lizhen and Liu, Wen and Xie, Zhenda and Liu, Yebin},
journal={arXiv preprint arXiv:2310.16818},
year={2023}
}