Repaint123: Fast and High-quality One Image to 3D Generation with Progressive Controllable 2D Repainting (ECCV2024)
Repaint123 crafts 3D content from a single image, matching 2D generation quality in just 2 minutes.
- Coarse stage: Gaussian Splatting optimized with SDS loss by Zero123 for geometry formation.
- Fine stage: Mesh optimized with MSE loss by Stable Diffusion for texture refinement.
- Stable Diffusion for high quality and controllable repainting for reference alignment --> view-consistent high-quality image generation.
- View-consistent high-quality images with simple MSE loss --> fast high-quality 3D content reconstruction.
Welcome to watch 👀 this repository for the latest updates.
✅ : Release project page. ✅ : Code release.
pip install -r requirements.txt
# a modified gaussian splatting (+ depth, alpha rendering)
git clone --recursive https://github.com/ashawkey/diff-gaussian-rasterization
pip install ./diff-gaussian-rasterization
# simple-knn
pip install ./simple-knn
# nvdiffrast
pip install git+https://github.com/NVlabs/nvdiffrast/
# kiuikit
pip install git+https://github.com/ashawkey/kiuikit
# To use MVdream, also install:
pip install git+https://github.com/bytedance/MVDream
# To use ImageDream, also install:
pip install git+https://github.com/bytedance/ImageDream/#subdirectory=extern/ImageDreamTested on:
- Ubuntu 22 with torch 1.12 & CUDA 11.6 on a V100.
- Windows 10 with torch 2.1 & CUDA 12.1 on a 3070.
Image-to-3D:
### preprocess
# background removal and recentering, save rgba at 256x256
python process.py data/name.jpg
# save at a larger resolution
python process.py data/name.jpg --size 512
# process all jpg images under a dir
python process.py data
### training gaussian stage
# train 500 iters (~1min) and export ckpt & coarse_mesh to logs
python main.py --config configs/image.yaml input=data/name_rgba.png save_path=name
# gui mode (supports visualizing training)
python main.py --config configs/image.yaml input=data/name_rgba.png save_path=name gui=True
# load and visualize a saved ckpt
python main.py --config configs/image.yaml load=logs/name_model.ply gui=True
# use an estimated elevation angle if image is not front-view (e.g., common looking-down image can use -30)
python main.py --config configs/image.yaml input=data/name_rgba.png save_path=name elevation=-30
### training mesh stage
# auto load coarse_mesh and refine 50 iters (~1min), export fine_mesh to logs
python main2.py --config configs/image.yaml input=data/name_rgba.png save_path=name
# specify coarse mesh path explicity
python main2.py --config configs/image.yaml input=data/name_rgba.png save_path=name mesh=logs/name_mesh.obj
# gui mode
python main2.py --config configs/image.yaml input=data/name_rgba.png save_path=name gui=True
# export glb instead of obj
python main2.py --config configs/image.yaml input=data/name_rgba.png save_path=name mesh_format=glb
### visualization
# gui for visualizing mesh
# `kire` is short for `python -m kiui.render`
kire logs/name.obj
# save 360 degree video of mesh (can run without gui)
kire logs/name.obj --save_video name.mp4 --wogui
# save 8 view images of mesh (can run without gui)
kire logs/name.obj --save images/name/ --wogui
### evaluation of CLIP-similarity
python -m kiui.cli.clip_sim data/name_rgba.png logs/name.objPlease check ./configs/image.yaml for more options.
This work is built on many amazing research works and open-source projects, thanks a lot to all the authors for sharing!
If you find our paper and code useful in your research, please consider giving a star ⭐ and citation 📝.
@inproceedings{zhang2024repaint123,
title={Repaint123: Fast and high-quality one image to 3d generation with progressive controllable repainting},
author={Zhang, Junwu and Tang, Zhenyu and Pang, Yatian and Cheng, Xinhua and Jin, Peng and Wei, Yida and Zhou, Xing and Ning, Munan and Yuan, Li},
booktitle={European Conference on Computer Vision},
pages={303--320},
year={2024},
organization={Springer}
}