CondiMen: Conditional Multi-Person Mesh Recovery

Romain Brégier, Fabien Baradel, Thomas Lucas, Salma Galaaoui, Matthieu Armando, Philippe Weinzaepfel, Grégory Rogez

CVPR'25 RHOBIN Workshop

A Bayesian method for multi-person human mesh recovery, modeling ambiguities in 3D pose and shape while enabling uncertainty handling and multi-view integration.

Installation

First, you need to clone the repo. The installation has been tested with python3.9 and CUDA 12.1.

We recommend to use a virtual enviroment for running CondiMen. Please run the following lines for creating the environment with venv:

python -m venv .condimen
source .condimen/bin/activate
pip install -r requirements.txt

Alternative environment using conda:

conda create -n condimen python=3.11
conda activate condimen
pip install -r requirements.txt

Download the demo checkpoint:

wget https://download.europe.naverlabs.com/condimen/condimen_checkpoint.pt condimen_checkpoint.pt

Besides these files, you also need to download the SMPLX model. You will need the neutral model to run the demo code. Please go to the corresponding website and register to get access to the downloads section. Download the model and place SMPLX_NEUTRAL.npz in ./models/smplx/.

Run CondiMen on images

The following command will run CondiMen on all images in the specified --img_folder, and save renderings of the reconstructions in --out_folder. The --model_name flag specifies the model to use.

python demo.py \
    --img_folder example_data \
    --out_folder demo_out \
    --checkpoint_path condimen_checkpoint.pt

License

The code is distributed under a non-commercial license.
See LICENSE and NOTICE for more information.

Citing

If you find this code useful for your research, please consider citing the following paper:

@inproceedings{condimen2025,
    title={CondiMen: Conditional Multi-Person Mesh Recovery},
    author={
            Br{\'e}gier, Romain and 
            Baradel, Fabien and 
            Lucas, Thomas and
            Galaaoui, Salma and 
            Armando, Matthieu and 
            Weinzaepfel, Philippe and 
            Rogez, Gr{\'e}gory
            },
    booktitle={CVPR RHOBIN Workshop},
    year={2025}
}