Home

Awesome

UV Volumes for Real-time Rendering of Editable Free-view Human Performance

Project Page | Paper | Latest arXiv | Supplementary

UV Volumes for Real-time Rendering of Editable Free-view Human Performance
Yue Chen*, Xuan Wang*, Xingyu Chen, Qi Zhang, Xiaoyu Li, Yu Guo†, Jue Wang, Fei Wang
(* equal contribution，† corresponding author)
CVPR 2023

UV Volumes for Real-time Rendering of Editable Free-view Human Performance

This repository is an official implementation of UV-Volumes using pytorch.

Installation

Please see INSTALL.md for manual installation.

Run the code on ZJU-MoCap

Please see INSTALL.md to download the dataset.

<!-- We provide the pretrained models at [here](https://). -->

Training on ZJU-MoCap

Take the training on sequence 313 as an example.

python3 train_net.py --cfg_file configs/zju_mocap_exp/313.yaml exp_name zju313 resume False output_depth True

You can monitor the training process by Tensorboard.

tensorboard --logdir data/record/UVvolume_ZJU

Test on ZJU-MoCap

Take the test on sequence 313 as an example.

<!-- 1. Download the corresponding pretrained model and put it to `$ROOT/data/trained_model/UVvolume_ZJU/zju313/latest.pth`. --> 
python3 run.py --type evaluate --cfg_file configs/zju_mocap_exp/313.yaml exp_name zju313 use_lpips True test.frame_sampler_interval 1 use_nb_mask_at_box True save_img True T_threshold 0.75

Run the code on CMU Panoptic

Please see INSTALL.md to download and process the dataset.

<!-- We provide the pretrained models at [here](https://). -->

Training on CMU Panoptic

Take the training on 171204_pose4_sample6 as an example.

python3 train_net.py --cfg_file configs/cmu_exp/p4s6.yaml exp_name p4s6 resume False output_depth True

You can monitor the training process by Tensorboard.

tensorboard --logdir data/record/UVvolume_CMU

Test on CMU Panoptic

Take the test on 171204_pose4_sample6 as an example.

<!-- 1. Download the corresponding pretrained model and put it to `$ROOT/data/trained_model/UVvolume_CMU/p4s6/latest.pth`. --> 
python3 run.py --type evaluate --cfg_file configs/cmu_exp/p4s6.yaml exp_name p4s6 use_lpips True test.frame_sampler_interval 1 use_nb_mask_at_box True save_img True

Citation

If you find this code useful for your research, please use the following BibTeX entry.

@inproceedings{chen2023uv,
  title={UV Volumes for real-time rendering of editable free-view human performance},
  author={Chen, Yue and Wang, Xuan and Chen, Xingyu and Zhang, Qi and Li, Xiaoyu and Guo, Yu and Wang, Jue and Wang, Fei},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={16621--16631},
  year={2023}
}

Acknowledge

Our code is based on the awesome pytorch implementation of NeuralBody. We appreciate all the contributors.