Populating 3D Scenes by Learning Human-Scene Interaction https://posa.is.tue.mpg.de/

Related tags

Deep LearningPOSA
Overview

Populating 3D Scenes by Learning Human-Scene Interaction

[Project Page] [Paper]

POSA Examples

License

Software Copyright License for non-commercial scientific research purposes. Please read carefully the following terms and conditions and any accompanying documentation before you download and/or use the POSA data, model and software, (the "Data & Software"), including 3D meshes, images, videos, textures, software, scripts, and animations. By downloading and/or using the Data & Software (including downloading, cloning, installing, and any other use of the corresponding github repository), you acknowledge that you have read these terms and conditions, understand them, and agree to be bound by them. If you do not agree with these terms and conditions, you must not download and/or use the Data & Software. Any infringement of the terms of this agreement will automatically terminate your rights under this License.

Description

This repository contains the training, random sampling, and scene population code used for the experiments in POSA.

Installation

To install the necessary dependencies run the following command:

    pip install -r requirements.txt

The code has been tested with Python 3.7, CUDA 10.0, CuDNN 7.5 and PyTorch 1.7 on Ubuntu 20.04.

Dependencies

POSA_dir

To be able to use the code you need to get the POSA_dir.zip. After unzipping, you should have a directory with the following structure:

POSA_dir
├── cam2world
├── data
├── mesh_ds
├── scenes
├── sdf
└── trained_models

The content of each folder is explained below:

  • trained_models contains two trained models. One is trained on the contact only and the other one is trained on contact and semantics.
  • data contains the train and test data extracted from the PROX Dataset and PROX-E Dataset.
  • scenes contains the 12 scenes from PROX Dataset
  • sdf contains the signed distance field for the scenes in the previous folder.
  • mesh_ds contains mesh downsampling and upsampling related files similar to the ones in COMA.

SMPL-X

You need to get the SMPLx Body Model. Please extract the folder and rename it to smplx_models and place it in the POSA_dir above.

AGORA

In addition, you need to get the POSA_rp_poses.zip file from AGORA Dataset and extract in the POSA_dir. This file contrains a number of test poses to be used in the next steps. Note that you don't need the whole AGORA dataset.

Finally run the following command or add it to your ~/.bashrc

export POSA_dir=Path of Your POSA_dir

Inference

You can test POSA using the trained models provided. Below we provide examples of how to generate POSA features and how to pupulate a 3D scene.

Random Sampling

To generate random features from a trained model, run the following command

python src/gen_rand_samples.py --config cfg_files/contact.yaml --checkpoint_path $POSA_dir/trained_models/contact.pt --pkl_file_path $POSA_dir/POSA_rp_poses/rp_aaron_posed_001_0_0.pkl --render 1 --viz 1 --num_rand_samples 3 

Or

python src/gen_rand_samples.py --config cfg_files/contact_semantics.yaml --checkpoint_path $POSA_dir/trained_models/contact_semantics.pt --pkl_file_path $POSA_dir/POSA_rp_poses/rp_aaron_posed_001_0_0.pkl --render 1 --viz 1 --num_rand_samples 3 

This will open a window showing the generated features for the specified pkl file. It also render the features to the folder random_samples in POSA_dir.

The number of generated feature maps can be controlled by the flag num_rand_samples.

If you don't have a screen, you can turn off the visualization --viz 0.

If you don't have CUDA installed then you can add this flag --use_cuda 0. This applies to all commands in this repository.

You can also run the same command on the whole folder of test poses

python src/gen_rand_samples.py --config cfg_files/contact_semantics.yaml --checkpoint_path $POSA_dir/trained_models/contact_semantics.pt --pkl_file_path $POSA_dir/POSA_rp_poses --render 1 --viz 1 --num_rand_samples 3 

Scene Population

Given a body mesh from the AGORA Dataset, POSA automatically places the body mesh in 3D scene.

python src/affordance.py --config cfg_files/contact_semantics.yaml --checkpoint_path $POSA_dir/trained_models/contact_semantics.pt --pkl_file_path $POSA_dir/POSA_rp_poses/rp_aaron_posed_001_0_0.pkl --scene_name MPH16 --render 1 --viz 1 

This will open a window showing the placed body in the scene. It also render the placements to the folder affordance in POSA_dir.

You can control the number of placements for the same body mesh in a scene using the flag num_rendered_samples, default value is 1.

The generated feature maps can be shown by setting adding --show_gen_sample 1

You can also run the same script on the whole folder of test poses

python src/affordance.py --config cfg_files/contact_semantics.yaml --checkpoint_path $POSA_dir/trained_models/contact_semantics.pt --pkl_file_path $POSA_dir/POSA_rp_poses --scene_name MPH16 --render 1 --viz 1 

To place clothed body meshes, you need to first buy the Renderpeople assets, or get the free models. Create a folder rp_clothed_meshes in POSA_dir and place all the clothed body .obj meshes in this folder. Then run this command:

python src/affordance.py --config cfg_files/contact_semantics.yaml --checkpoint_path $POSA_dir/trained_models/contact_semantics.pt --pkl_file_path $POSA_dir/POSA_rp_poses/rp_aaron_posed_001_0_0.pkl --scene_name MPH16 --render 1 --viz 1 --use_clothed_mesh 1

Testing on Your Own Poses

POSA has been tested on the AGORA dataset only. Nonetheless, you can try POSA with any SMPL-X poses you have. You just need a .pkl file with the SMPLX body parameters and the gender. Your SMPL-X vertices must be brought to a canonical form similar to the POSA training data. This means the vertices should be centered at the pelvis joint, the x axis pointing to the left, the y axis pointing backward, and the z axis pointing upwards. As shown in the figure below. The x,y,z axes are denoted by the red, green, blue colors respectively.

canonical_form

See the function pkl_to_canonical in data_utils.py for an example of how to do this transformation.

Training

To retrain POSA from scratch run the following command

python src/train_posa.py --config cfg_files/contact_semantics.yaml

Visualize Ground Truth Data

You can also visualize the training data

python src/show_gt.py --config cfg_files/contact_semantics.yaml --train_data 1

Or test data

python src/show_gt.py --config cfg_files/contact_semantics.yaml --train_data 0

Note that the ground truth data has been downsampled to speed up training as explained in the paper. See training details in appendices.

Citation

If you find this Model & Software useful in your research we would kindly ask you to cite:

@inproceedings{Hassan:CVPR:2021,
    title = {Populating {3D} Scenes by Learning Human-Scene Interaction},
    author = {Hassan, Mohamed and Ghosh, Partha and Tesch, Joachim and Tzionas, Dimitrios and Black, Michael J.},
    booktitle = {Proceedings {IEEE/CVF} Conf.~on Computer Vision and Pattern Recognition ({CVPR})},
    month = jun,
    month_numeric = {6},
    year = {2021}
}

If you use the extracted training data, scenes or sdf the please cite:

@inproceedings{PROX:2019,
  title = {Resolving {3D} Human Pose Ambiguities with {3D} Scene Constraints},
  author = {Hassan, Mohamed and Choutas, Vasileios and Tzionas, Dimitrios and Black, Michael J.},
  booktitle = {International Conference on Computer Vision},
  month = oct,
  year = {2019},
  url = {https://prox.is.tue.mpg.de},
  month_numeric = {10}
}
@inproceedings{PSI:2019,
  title = {Generating 3D People in Scenes without People},
  author = {Zhang, Yan and Hassan, Mohamed and Neumann, Heiko and Black, Michael J. and Tang, Siyu},
  booktitle = {Computer Vision and Pattern Recognition (CVPR)},
  month = jun,
  year = {2020},
  url = {https://arxiv.org/abs/1912.02923},
  month_numeric = {6}
}

If you use the AGORA test poses, the please cite:

@inproceedings{Patel:CVPR:2021,
  title = {{AGORA}: Avatars in Geography Optimized for Regression Analysis},
  author = {Patel, Priyanka and Huang, Chun-Hao P. and Tesch, Joachim and Hoffmann, David T. and Tripathi, Shashank and Black, Michael J.},
  booktitle = {Proceedings IEEE/CVF Conf.~on Computer Vision and Pattern Recognition (CVPR)},
  month = jun,
  year = {2021},
  month_numeric = {6}
}

Contact

For commercial licensing (and all related questions for business applications), please contact [email protected].

Owner
Mohamed Hassan
Mohamed Hassan
Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking

Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking We revisit and address issues with Oxford 5k and Paris 6k image retrieval benchm

Filip Radenovic 188 Dec 17, 2022
PyTorch implementation of the wavelet analysis from Torrence & Compo

Continuous Wavelet Transforms in PyTorch This is a PyTorch implementation for the wavelet analysis outlined in Torrence and Compo (BAMS, 1998). The co

Tom Runia 262 Dec 21, 2022
Official PyTorch implementation for paper Context Matters: Graph-based Self-supervised Representation Learning for Medical Images

Context Matters: Graph-based Self-supervised Representation Learning for Medical Images Official PyTorch implementation for paper Context Matters: Gra

49 Nov 23, 2022
A web application that provides real time temperature and humidity readings of a house.

About A web application which provides real time temperature and humidity readings of a house. If you're interested in the data collected so far click

Ben Thompson 3 Jan 28, 2022
Official PyTorch implementation of "RMGN: A Regional Mask Guided Network for Parser-free Virtual Try-on" (IJCAI-ECAI 2022)

RMGN-VITON RMGN: A Regional Mask Guided Network for Parser-free Virtual Try-on In IJCAI-ECAI 2022(short oral). [Paper] [Supplementary Material] Abstra

27 Dec 01, 2022
Emotional conditioned music generation using transformer-based model.

This is the official repository of EMOPIA: A Multi-Modal Pop Piano Dataset For Emotion Recognition and Emotion-based Music Generation. The paper has b

hung anna 96 Nov 09, 2022
👐OpenHands : Making Sign Language Recognition Accessible (WiP 🚧👷‍♂️🏗)

👐 OpenHands: Sign Language Recognition Library Making Sign Language Recognition Accessible Check the documentation on how to use the library: ReadThe

AI4Bhārat 69 Dec 12, 2022
Code for the paper A Theoretical Analysis of the Repetition Problem in Text Generation

A Theoretical Analysis of the Repetition Problem in Text Generation This repository share the code for the paper "A Theoretical Analysis of the Repeti

Zihao Fu 37 Nov 21, 2022
"3D Human Texture Estimation from a Single Image with Transformers", ICCV 2021

Texformer: 3D Human Texture Estimation from a Single Image with Transformers This is the official implementation of "3D Human Texture Estimation from

XiangyuXu 193 Dec 05, 2022
An executor that loads ONNX models and embeds documents using the ONNX runtime.

ONNXEncoder An executor that loads ONNX models and embeds documents using the ONNX runtime. Usage via Docker image (recommended) from jina import Flow

Jina AI 2 Mar 15, 2022
Viewmaker Networks: Learning Views for Unsupervised Representation Learning

Viewmaker Networks: Learning Views for Unsupervised Representation Learning Alex Tamkin, Mike Wu, and Noah Goodman Paper link: https://arxiv.org/abs/2

Alex Tamkin 31 Dec 01, 2022
Code for the paper "Zero-shot Natural Language Video Localization" (ICCV2021, Oral).

Zero-shot Natural Language Video Localization (ZSNLVL) by Pseudo-Supervised Video Localization (PSVL) This repository is for Zero-shot Natural Languag

Computer Vision Lab. @ GIST 37 Dec 27, 2022
This is the source code for the experiments related to the paper Unsupervised Audio Source Separation Using Differentiable Parametric Source Models

Unsupervised Audio Source Separation Using Differentiable Parametric Source Models This is the source code for the experiments related to the paper Un

30 Oct 19, 2022
Joint deep network for feature line detection and description

SOLD² - Self-supervised Occlusion-aware Line Description and Detection This repository contains the implementation of the paper: SOLD² : Self-supervis

Computer Vision and Geometry Lab 427 Dec 27, 2022
StarGAN - Official PyTorch Implementation (CVPR 2018)

StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation

Yunjey Choi 5.1k Dec 30, 2022
Created as part of CS50 AI's coursework. This AI makes use of knowledge entailment to calculate the best probabilities to win Minesweeper.

Minesweeper-AI Created as part of CS50 AI's coursework. This AI makes use of knowledge entailment to calculate the best probabilities to win Minesweep

Beckham 0 Jul 20, 2022
Optimizing synthesizer parameters using gradient approximation

Optimizing synthesizer parameters using gradient approximation NASH 2021 Hackathon! These are some experiments I conducted during NASH 2021, the Neura

Jordie Shier 10 Feb 10, 2022
Modifications of the official PyTorch implementation of StyleGAN3. Let's easily generate images and videos with StyleGAN2/2-ADA/3!

Alias-Free Generative Adversarial Networks (StyleGAN3) Official PyTorch implementation of the NeurIPS 2021 paper Alias-Free Generative Adversarial Net

Diego Porres 185 Dec 24, 2022
Callable PyTrees and filtered JIT/grad transformations => neural networks in JAX.

Equinox Callable PyTrees and filtered JIT/grad transformations = neural networks in JAX Equinox brings more power to your model building in JAX. Repr

Patrick Kidger 909 Dec 30, 2022
A modular domain adaptation library written in PyTorch.

A modular domain adaptation library written in PyTorch.

Kevin Musgrave 225 Dec 29, 2022