ECON: Explicit Clothed humans Optimized via Normal integration

Yuliang Xiu · Jinlong Yang · Xu Cao · Dimitrios Tzionas · Michael J. Black

CVPR 2023 (Highlight)

ECON is designed for "Human digitization from a color image", which combines the best properties of implicit and explicit representations, to infer high-fidelity 3D clothed humans from in-the-wild images, even with loose clothing or in challenging poses. ECON also supports multi-person reconstruction and SMPL-X based animation.

HuggingFace Demo	Google Colab	Blender Add-on	Windows	Docker

Applications


"3D guidance" for SHHQ Dataset	multi-person reconstruction w/ occlusion

"All-in-One" Blender add-on	SMPL-X based Animation (Instruction)

News 🚩

[2023/06/01] Lee Kwan Joong updates a Blender Addon (Github, Tutorial).
[2023/04/16] is ready to use!
[2023/02/27] ECON got accepted by CVPR 2023 as Highlight (top 10%)!
[2023/01/12] Carlos Barreto creates a Blender Addon (Download, Tutorial).
[2023/01/08] Teddy Huang creates install-with-docker for ECON .
[2023/01/06] Justin John and Carlos Barreto creates install-on-windows for ECON .
[2022/12/22] is now available, created by Aron Arzoomand.
[2022/12/15] Both demo and arXiv are available.

Key idea: d-BiNI

d-BiNI jointly optimizes front-back 2.5D surfaces such that: (1) high-frequency surface details agree with normal maps, (2) low-frequency surface variations, including discontinuities, align with SMPL-X surfaces, and (3) front-back 2.5D surface silhouettes are coherent with each other.

Front-view	Back-view	Side-view

Please consider cite BiNI if it also helps on your project

@inproceedings{cao2022bilateral,
  title={Bilateral normal integration},
  author={Cao, Xu and Santo, Hiroaki and Shi, Boxin and Okura, Fumio and Matsushita, Yasuyuki},
  booktitle={Computer Vision--ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23--27, 2022, Proceedings, Part I},
  pages={552--567},
  year={2022},
  organization={Springer}
}

Table of Contents

Instructions
Demos
Citation

Instructions

See installion doc for Docker to run a docker container with pre-built image for ECON demo
See installion doc for Windows to install all the required packages and setup the models on Windows
See installion doc for Ubuntu to install all the required packages and setup the models on Ubuntu
See magic tricks to know a few technical tricks to further improve and accelerate ECON
See testing to prepare the testing data and evaluate ECON

Demos

Quick Start

# For single-person image-based reconstruction (w/ l visualization steps, 1.8min)
python -m apps.infer -cfg ./configs/econ.yaml -in_dir ./examples -out_dir ./results

# For multi-person image-based reconstruction (see config/econ.yaml)
python -m apps.infer -cfg ./configs/econ.yaml -in_dir ./examples -out_dir ./results -multi

# To generate the demo video of reconstruction results
python -m apps.multi_render -n <file_name>

Animation with SMPL-X sequences (ECON + HybrIK-X)

# 1. Use HybrIK-X to estimate SMPL-X pose sequences from input video
# 2. Rig ECON's reconstruction mesh, to be compatible with SMPL-X's parametrization (-dress for dress/skirts).
# 3. Animate with SMPL-X pose sequences obtained from HybrIK-X, getting <file_name>_motion.npz
# 4. Render the frames with Blender (rgb-partial texture, normal-normal colors), and combine them to get final video

python -m apps.avatarizer -n <file_name>
python -m apps.animation -n <file_name> -m <motion_name>

# Note: to install missing python packages into Blender
# blender -b --python-expr "__import__('pip._internal')._internal.main(['install', 'moviepy'])"

wget https://download.is.tue.mpg.de/icon/econ_empty.blend
blender -b --python apps.blender_dance.py -- normal <file_name> 10 > /tmp/NULL

Please consider cite HybrIK-X if it also helps on your project

@article{li2023hybrik,
  title={HybrIK-X: Hybrid Analytical-Neural Inverse Kinematics for Whole-body Mesh Recovery},
  author={Li, Jiefeng and Bian, Siyuan and Xu, Chao and Chen, Zhicun and Yang, Lixin and Lu, Cewu},
  journal={arXiv preprint arXiv:2304.05690},
  year={2023}
}

Gradio Demo

We also provide a UI for testing our method that is built with gradio. This demo also supports pose&prompt guided human image generation! Running the following command in a terminal will launch the demo:

git checkout main
python app.py

This demo is also hosted on HuggingFace Space

Full Texture Generation

Please firstly follow the TEXTure's installation to setup the env of TEXTure.

# generate required UV atlas
python -m apps.avatarizer -n <file_name> -uv

# generate new texture using TEXTure
git clone https://github.com/YuliangXiu/TEXTure
cd TEXTure
ln -s ../ECON/results/econ/cache
python -m scripts.run_texture --config_path=configs/text_guided/avatar.yaml

Then check ./experiments/<file_name>/mesh for the results.

Please consider cite TEXTure if it also helps on your project

@article{richardson2023texture,
  title={Texture: Text-guided texturing of 3d shapes},
  author={Richardson, Elad and Metzer, Gal and Alaluf, Yuval and Giryes, Raja and Cohen-Or, Daniel},
  journal={ACM Transactions on Graphics (TOG)},
  publisher={ACM New York, NY, USA},
  year={2023}
}

More Qualitative Results


Challenging Poses

Loose Clothes

Citation

@inproceedings{xiu2023econ,
  title     = {{ECON: Explicit Clothed humans Optimized via Normal integration}},
  author    = {Xiu, Yuliang and Yang, Jinlong and Cao, Xu and Tzionas, Dimitrios and Black, Michael J.},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month     = {June},
  year      = {2023},
}

Acknowledgments

We thank Lea Hering and Radek Daněček for proof reading, Yao Feng, Haven Feng, and Weiyang Liu for their feedback and discussions, Tsvetelina Alexiadis for her help with the AMT perceptual study.

Here are some great resources we benefit from:

ICON for SMPL-X Body Fitting
BiNI for Bilateral Normal Integration
MonoPortDataset for Data Processing, MonoPort for fast implicit surface query
rembg for Human Segmentation
MediaPipe for full-body landmark estimation
PyTorch-NICP for non-rigid registration
smplx, PyMAF-X, PIXIE for Human Pose & Shape Estimation
CAPE and THuman for Dataset
PyTorch3D for Differential Rendering

Some images used in the qualitative examples come from pinterest.com.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No.860768 (CLIPE Project).

Contributors

Kudos to all of our amazing contributors! ECON thrives through open-source. In that spirit, we welcome all kinds of contributions from the community.

Contributor avatars are randomly shuffled.

License

This code and model are available for non-commercial scientific research purposes as defined in the LICENSE file. By downloading and using the code and model you agree to the terms in the LICENSE.

Disclosure

MJB has received research gift funds from Adobe, Intel, Nvidia, Meta/Facebook, and Amazon. MJB has financial interests in Amazon, Datagen Technologies, and Meshcapade GmbH. While MJB is a part-time employee of Meshcapade, his research was performed solely at, and funded solely by, the Max Planck Society.

Contact

For technical questions, please contact yuliang.xiu@tue.mpg.de

For commercial licensing, please contact ps-licensing@tue.mpg.de

ouceduxzk / ECON