NIKI: Neural Inverse Kinematics with Invertible Neural Networks for 3D HPS
This repo contains the code of our paper:
NIKI: Neural Inverse Kinematics with Invertible Neural Networks for 3D Human Pose and Shape Estimation
Jiefeng Li*, Siyuan Bian*, Qi Liu, Jiasheng Tang, Fan Wang, Cewu Lu
In CVPR 2023
Different Inverse Kinematics Framework
Installation instructions
# 1. Create a conda virtual environment.
conda create -n niki python=3.8 -y
conda activate niki
# 2. Install PyTorch
conda install pytorch==1.13.0 torchvision==0.14.0 torchaudio==0.13.0 pytorch-cuda=11.6 -c pytorch -c nvidia
# 3. Install PyTorch3D (Optional, only for visualization)
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub
pip install git+ssh://[email protected]/facebookresearch/pytorch3d.git
pip install -r requirements.txt
# 4. Pull our code
git clone [email protected]:Jeff-sjtu/NIKI.git
cd NIKIModels
- Download our pretrained model from [Google Drive | Baidu (code:
z4iv)].
Fetch data
- Download parsed annotations from [Google Drive].
Train
python scripts/train.py --cfg configs/NIKI-ts.yaml --exp-id niki-tsEvaluation
python scripts/validate.py --cfg configs/NIKI-ts.yaml --ckpt niki-ts.pthDemo
Download pretrained HybrIK and Single-stage NIKI models from onedrive link, and put them in exp/ folder.
python scripts/demo.py --video-name {VIDEO-PATH} -out-dir {OUTPUT-DIR}Citing
If our code helps your research, please consider citing the following paper:
@inproceedings{li2023niki,
title = {{NIKI}: Neural Inverse Kinematics with Invertible Neural Networks for 3D Human Pose and Shape Estimation},
author = {Li, Jiefeng and Bian, Siyuan and Liu, Qi and Tang, Jiasheng and Wang, Fan and Lu, Cewu},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
}