Awesome

OpenMPL

MPL: Lifting 3D Human Pose from Multi-view 2D Poses, T-CAP ECCV24

Seyed Abolfazl Ghasemzadeh, Alexandre Alahi, Christophe De Vleeschouwer

arxiv 2408.10805

Installation

To get started with OpenMPL, follow these steps:

Steps

Clone the Repository:

git clone git@github.com:aghasemzadeh/OpenMPL.git
cd OpenMPL

Install Dependencies:
```
pip install -r requirements.txt
```
Install amass framework through the following link to be able to run MHP:

https://github.com/nghorbani/amass

Install mmpose for running off-the-shelf 2D pose estimation:

https://github.com/open-mmlab/mmpose
Datasets Setup:

You will need AMASS dataset for training (+ camera calibrations from your test dataset) and a dataset for testing in MPL. We test our framework on two datasets: CMU and Human3.6M.

CMU

For CMU data, please follow panoptic-toolbox to prepare images and annotations. Then run the command below after correcting the paths:
```
cd MPL/data
sh preprocess_cmu_panoptic_all_cams.sh
```
Human3.6M

For Human36M data, please follow H36M-Toolbox to prepare images and annotations. Then run the command below after correcting the paths:
```
cd MPL/data
sh preprocess_h36m.sh
```
AMASS

For running MHP (generating 3D dataset for MPL), you need to download the amass dataset. Please follow AMASS for downloading the pose SMPL data. For running MHP, you will need to install amass framework and follow their guidelines.

After correcting the paths, you can run the command below:
```
cd MPH
sh run_mmpose_00_cmu_calibs.sh      # for CMU
sh run_mmpose_00_h36m_calibs.sh     # for Human3.6M
```
Remember to go through the code in these files to be sure you run everything. There are some parallel scripts that you need to take care yourself. (Some are commented)

Citation

Please make sure to cite our paper if you utilize our code.

@ARTICLE{Ghasemzadeh2024-ln,
  title         = "{MPL}: Lifting {3D} Human Pose from Multi-view {2D} Poses",
  author        = "Ghasemzadeh, Seyed Abolfazl and Alahi, Alexandre and De
                   Vleeschouwer, Christophe",
  month         =  aug,
  year          =  2024,
  copyright     = "http://creativecommons.org/licenses/by-nc-sa/4.0/",
  archivePrefix = "arXiv",
  primaryClass  = "cs.CV",
  eprint        = "2408.10805"
}

Acknowledgement

Part of our code is borrowed from PPT and PoseFormer. We thank the authors for releasing the codes.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Happy coding! If you have any questions, feel free to open an issue or contact us directly.