ICON: Implicit Clothed humans Obtained from Normals

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

CVPR 2022

News 🚩

[2022/07/30] are both available.
[2022/07/26] New cloth-refinement module is released, try -loop_cloth.
[2022/06/13] ETH Zürich students from 3DV course create an add-on for garment-extraction.
[2022/05/16] BEV is supported as optional HPS by Yu Sun, see commit #060e265.
[2022/05/15] Training code is released, please check Training Instruction.
[2022/04/26] HybrIK (SMPL) is supported as optional HPS by Jiefeng Li, see commit #3663704.
[2022/03/05] PIXIE (SMPL-X), PARE (SMPL), PyMAF (SMPL) are all supported as optional HPS.

Table of Contents

Who needs ICON
Installation
Dataset
Training
Evaluation
Add-on
Demo
Citation
Acknowledgments
License
Disclosure
Contact

Who needs ICON?

If you want to Train & Evaluate on PIFu / PaMIR / ICON using your own data, please check dataset.md to prepare dataset, training.md for training, and evaluation.md for benchmark evaluation.
Given a raw RGB image, you could get:
- image (png):
  - segmented human RGB
  - normal maps of body and cloth
  - pixel-aligned normal-RGB overlap
- mesh (obj):
  - SMPL-(X) body from PyMAF, PIXIE, PARE, HybrIK, BEV
  - 3D clothed human reconstruction
  - 3D garments (requires 2D mask)
- video (mp4):
  - self-rotated clothed human


ICON's intermediate results

ICON's SMPL Pose Refinement

Image -- overlapped normal prediction -- ICON -- refined ICON

3D Garment extracted from ICON using 2D mask

Installation

Please follow the Installation Instruction to setup all the required packages, extra data, and models.

Dataset

Please follow the Dataset Instruction to generate the train/val/test dataset from THuman2.0

Training

Please follow the Training Instruction to train your own model using THuman2.0

Evaluation

Please follow the Evaluation Instruction to benchmark trained models on THuman2.0

Add-on

Garment Extraction from Fashion Images, supported by ETH Zürich students as 3DV course project.

Demo

cd ICON

# PIFu* (*: re-implementation)
python -m apps.infer -cfg ./configs/pifu.yaml -gpu 0 -in_dir ./examples -out_dir ./results -loop_cloth 100 -export_video

# PaMIR* (*: re-implementation)
python -m apps.infer -cfg ./configs/pamir.yaml -gpu 0 -in_dir ./examples -out_dir ./results -loop_smpl 100 -loop_cloth 200 -export_video

# ICON w/ global filter (better visual details --> lower Normal Error))
python -m apps.infer -cfg ./configs/icon-filter.yaml -gpu 0 -in_dir ./examples -out_dir ./results -hps_type {pymaf/pixie/pare/hybrik/bev} -loop_smpl 100 -loop_cloth 200 -export_video

# ICON w/o global filter (higher evaluation scores --> lower P2S/Chamfer Error))
python -m apps.infer -cfg ./configs/icon-nofilter.yaml -gpu 0 -in_dir ./examples -out_dir ./results -hps_type {pymaf/pixie/pare/hybrik/bev} -loop_smpl 100 -loop_cloth 200 -export_video

More Qualitative Results


Comparison with other state-of-the-art methods

Predicted normals on in-the-wild images with extreme poses

Citation

@inproceedings{xiu2022icon,
  title     = {{ICON}: {I}mplicit {C}lothed humans {O}btained from {N}ormals},
  author    = {Xiu, Yuliang and Yang, Jinlong and Tzionas, Dimitrios and Black, Michael J.},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month     = {June},
  year      = {2022},
  pages     = {13296-13306}
}

Acknowledgments

We thank Yao Feng, Soubhik Sanyal, Qianli Ma, Xu Chen, Hongwei Yi, Chun-Hao Paul Huang, and Weiyang Liu for their feedback and discussions, Tsvetelina Alexiadis for her help with the AMT perceptual study, Taylor McConnell for her voice over, Benjamin Pellkofer for webpage, and Yuanlu Xu's help in comparing with ARCH and ARCH++.

Special thanks to Vassilis Choutas for sharing the code of bvh-distance-queries

Here are some great resources we benefit from:

MonoPortDataset for Data Processing
PaMIR, PIFu, PIFuHD, and MonoPort for Benchmark
SCANimate and AIST++ for Animation
rembg for Human Segmentation
PyTorch-NICP for normal-based non-rigid refinement
smplx, PARE, PyMAF, PIXIE, BEV, and HybrIK 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).

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 was a part-time employee of Amazon during this project, his research was performed solely at, and funded solely by, the Max Planck Society.

Contact

For more questions, please contact icon@tue.mpg.de

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

gyhandy/ICON