Deep Geometry Analysis for Uncalibrated Multi-View 3D Human Pose Estimation

This repository is the official implementation of [Deep Geometry Analysis for Uncalibrated Multi-View 3D Human Pose Estimation].

Requirements

To install requirements:

#1. Create a conda virtual environment.
conda create -n mvhpe python=3.7.11
conda activate mvhpe

#2. Install Pytorch
pip install torch==1.8.1+cu101 torchvision==0.9.1+cu101 torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html

#3. Install requirements.
pip install -r requirements.txt

Preparing Data and Rre-trained model

1. Download the required data.

   • Download our data from Google Drive
   • Download our pretrained model from Google Drive

2. You need to add the data and checkpoint as below.

|-- data
`-- |-- h36m_sub1.npz
    `-- ...
    `-- h36m_sub11.npz
    `-- score.pkl
|-- checkpoint
`-- |-- h36m_cpn_uncalibration.pth
    `-- h36m_gt_uncalibration.pth
    `-- h36m_cpn_calibration.pth

Evaluation

To evaluate our model, run:

python eval_h36m_gt_uncalibration.py --test --in_chans 2 --previous_dir ./checkpoint/h36m_gt_uncalibration.pth --root_path /home/zzj/TMM/MV-3D-HPE/data/ --gpu 0
python eval_h36m_cpn_uncalibration.py --test --out_chans 3 --previous_dir ./checkpoint/h36m_cpn_uncalibration.pth --root_path /home/zzj/TMM/MV-3D-HPE/data/ --gpu 0
python eval_h36m_cpn_calibration.py --test --out_chans 2 --previous_dir ./checkpoint/h36m_cpn_calibration.pth --root_path /home/zzj/TMM/MV-3D-HPE/data/ --gpu 0

Results

Our model achieves the following performance on Human3.6M:

Methods	Camera	MPJPE
Ours (GT, T=27)	Uncalibration	5.2mm
Ours (CPN, T=27)	Uncalibration	24.5mm
Ours (CPN, T=27)	Calibration	23.8mm

ToDos:

Implement the functions to

• We will provide training code after the paper is accepted.

License

By downloading and using this code you agree to the terms in the LICENSE. Third-party datasets and software are subject to their respective licenses.