saic-vul / point_based_clothing

Official PyTorch code for the paper: "Point-Based Modeling of Human Clothing" (ICCV 2021)

Point-Based Modeling of Human Clothing

Paper | Project page | Video

This is an official PyTorch code repository of the paper "Point-Based Modeling of Human Clothing" (accepted to ICCV, 2021).

Setup

Build docker

• Prerequisites: your nvidia driver should support cuda 10.2, Windows or Mac are not supported.
• Clone repo:
  • git clone https://github.com/izakharkin/point_based_clothing.git
  • cd point_based_clothing
  • git submodule init && git submodule update
• Docker setup:
  • Install docker engine
  • Install nvidia-docker
  • Set nvidia your default runtime for docker
  • Make docker run without sudo: create docker group and add current user to it:

    sudo groupadd docker
    sudo usermod -aG docker $USER
    

  • Reboot
• Download 10_nvidia.json and place it in the docker/ folder
• Create docker image:
  • Build on your own: run 2 commands
• Inside the docker container: source activate pbc

Download data

• Download the SMPL neutral model from SMPLify project page:
  • Register, go to the Downloads section, download SMPLIFY_CODE_V2.ZIP, and unpack it;
  • Move smplify_public/code/models/basicModel_neutral_lbs_10_207_0_v1.0.0.pkl to data/smpl_models/SMPL_NEUTRAL.pkl.
• Download models checkpoints (~570 Mb): Google Drive and place them to the checkpoints/ folder;
• Download a sample data we provide to check the appearance fitting (~480 Mb): Google Drive, unpack it, and place psp/ folder to the samples/ folder.

Custom data

To run our pipeline on custom data (images or videos):

• run our fork of Graphonomy to obtain clothing segmentation mask in our format;
• run e.g. SMPLify or any other suitable method to obtain the SMPL-parameters (3D body pose and shape ground truth).

We recommend to run these methods on internet_images/ test dataset to make sure that your outputs exactly match the format inside internet_images/segmentations/cloth and internet_images/smpl/results.

Run

We provide scripts for geometry fitting and inference and appearance fitting and inference.

Geometry (outfit code)

Fitting

To fit a style outfit code to a single image one can run:

python fit_outfit_code.py --config_name=outfit_code/psp

The learned outfit codes are saved to out/outfit_code/outfit_codes_<dset_name>.pkl by default. The visualization of the process is in out/outfit_code/vis_<dset_name>/:

• Coarse fitting stage: four outfit codes initialized randomly and being optimized simultaneosly.

• Fine fitting stage: mean of found outfit codes is being optimized further to possibly imrove the reconstruction.

Note: visibility_thr hyperparameter in fit_outfit_code.py may affect the quality of result point cloud (e.f. make it more sparse). Feel free to tune it if the result seems not perfect.

Inference

To further infer the fitted outfit style on the train or on new subjects please see infer_outfit_code.ipynb. To run jupyter notebook server from the docker, run this inside the container:

jupyter notebook --ip=0.0.0.0 --port=8087 --no-browser 

Appearance (neural descriptors)

Fitting

To fit a clothing appearance to a sequence of frames one can run:

python fit_appearance.py --config_name=appearance/psp_male-3-casual

The learned neural descriptors ntex0_<epoch>.pth and neural rendering network weights model0_<epoch>.pth are saved to out/appearance/<dset_name>/<subject_id>/<experiment_dir>/checkpoints/ by default. The visualization of the process is in out/appearance/<dset_name>/<subject_id>/<experiment_dir>/visuals/.

Inference

To further infer the fitted clothing point cloud and its appearance on the train or on new subjects please see infer_appearance.ipynb. To run jupyter notebook server from the docker, run this inside the container:

jupyter notebook --ip=0.0.0.0 --port=8087 --no-browser 

Citation

If you find our work helpful, please do not hesitate to cite us:

@InProceedings{Zakharkin_2021_ICCV,
    author    = {Zakharkin, Ilya and Mazur, Kirill and Grigorev, Artur and Lempitsky, Victor},
    title     = {Point-Based Modeling of Human Clothing},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {14718-14727}
}

Non-commercial use only.

Related projects

• Cloud Transformers
• Neural Point-Based Graphics
• Neural Textures and StylePeople

We also thank the authors of Cloth3D and PeopleSnapshot datasets.