Awesome

Official Implementation of the paper "Implicit Neural Representations for Variable Length Human Motion Generation" (ECCV 2022)

Main figure

Bibtex

Please consider citing this work, if you find this code useful.

@article{cervantes2022implicit,
  title={Implicit Neural Representations for Variable Length Human Motion Generation},
  author={Cervantes, Pablo and Sekikawa, Yusuke and Sato, Ikuro and Shinoda, Koichi},
  journal={arXiv preprint arXiv:2203.13694},
  year={2022}
}

Installation

pip install -r requirements.txt
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

To install Pytorch3D follow the instructions here.

For CUDA builds with versions earlier than CUDA 11, set CUB_HOME

pip install "git+https://github.com/facebookresearch/pytorch3d.git@v0.3.0"

Data Preparation

Download datasets

Download evaluation models

Download SMPL model

Usage

Training

python3 ./ImplicitMotion/main.py 
	--path_config=/path/to/config-file

Before training you need to prepare a configuration file. Configurations for the experiments in the paper are provided here. Modifications for the following keyword arguments are necessary:

path_dataset: Path to dataset folder
path_results_base: Path to folder for saving checkpoints, etc. (arbitrary).
path_smpl: Path to SMPL file (.pkl)

Evaluation

python3 ./ImplicitMotion/test/test_metric.py 
	--path_results /path/to/results
	--path_classifier /path/to/classifier
	--variable_length_testing
	--metrics

Visualization

python3 ./ImplicitMotion/test/test_metric.py
	--path_results /path/to/results
	--path_classifier /path/to/classifier
	--variable_length_testing
	--video
	--num_videos 1
	--video_length 60

License

This code is distributed under an MIT LICENSE.