HumanRef

Jingbo Zhang, Xiaoyu Li, Qi Zhang, Yanpei Cao, Ying Shan, Jing Liao

| Project Page | Paper |

This repository contains the official implementation of the paper "HumanRef: Single Image to 3D Human Generation via Reference-Guided Diffusion". Note that, this code is forked from threestudio for 3D representation and rendering pipeline.

Installation

conda env create -f environment.yml
conda activate humanref

pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch
pip install git+https://github.com/openai/CLIP.git
pip install git+https://github.com/huggingface/diffusers.git@ce5504934ac484fca39a1a5434ecfae09eabdf41

git clone https://github.com/NVlabs/nvdiffrast
cd nvdiffrast && pip install .
cd .. && rm -rf nvdiffrast

Download pre-trained models

1. download econ_weights

mkdir -p Weights/econ_weights && cd Weights
sh ./third_parties/ECON/fetch_data.sh
mv data/* econ_weights/
cd ..

2. download 'blip2-opt-2.7b' & 'stable-diffusion-v1-5'

cd Weights
git lfs install
git clone https://huggingface.co/Salesforce/blip2-opt-2.7b
git clone https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5
cd ..

3. download weights of LPIPS

mkdir -p Weights/LPIPS && cd Weights/LPIPS
wget https://download.pytorch.org/models/vgg16-397923af.pth
cd ../..

4. download weights of SCHP

mkdir -p Weights/SCHP

download exp-schp-201908261155-lip.pth and put it in 'Weights/SCHP'

Quickstart

Preprocessing: Estimate the SMPL-X body mesh using ECON and segment the input image into RGBA format

You need to modify the path setting ('os.environ['WEIGHT_PATH']') in 'third_parties/ECON/run_ECON_smpl.py'.

python third_parties/ECON/run_ECON_smpl.py --in_dir='./data/image_000355.jpg' --out_dir='./data/Results_ECON'

Optimization

python run.py --config configs/humanref.yaml --train --gpu 0 image_path='./data/Results_ECON/image_000355/econ/imgs_crop/image_000355_0_rgba.png'

Citation

If you find our code or paper helps, please consider citing:

@inproceedings{zhang2024humanref,
  title={Humanref: Single image to 3d human generation via reference-guided diffusion},
  author={Zhang, Jingbo and Li, Xiaoyu and Zhang, Qi and Cao, Yanpei and Shan, Ying and Liao, Jing},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={1844--1854},
  year={2024}
}