Awesome
EventHPE: Event-based 3D Human Pose and Shape Estimation
Shihao Zou, Chuan Guo, Xinxin Zuo, Sen Wang, Xiaoqin Hu, Shoushun Chen, Minglun Gong and Li Cheng. ICCV 2021.
Dataset
You can download the data from Google Drive or Microsoft OneDrive, which consists of
- preprocessed data
- events_256 (event frames converted from raw events data, resolution 256x256)
- full_pic_256 (gray-scale images)
- pose_events (annotated poses of gray-scale images)
- hmr_results (inferred poses of gray-scale images using HMR)
- vibe_results_0802 (inferred poses of gray-scale images using VIBE)
- pred_flow_events_256 (inferred optical flow from event frames)
- model (train/test on a snippet of 8 frames)
- raw events data (Please contact Shihao Zou szou2@ualberta.ca for the access.)
Requirements
python 3.7.5
torch 1.7.0
opendr 0.78 (for render SMPL shape, installed successfully only under ubuntu 18.04)
cv2 4.1.1
To download the SMPL model go to this project website and register to get access to the downloads section. Place under ./smpl_model. The model version used in our project is
basicModel_m_lbs_10_207_0_v1.0.0.pkl
basicModel_neutral_lbs_10_207_0_v1.0.0.pkl
Citation
If you would like to use our code or dataset, please cite either
@inproceedings{zou2021eventhpe,
title={EventHPE: Event-based 3D Human Pose and Shape Estimation},
author={Zou, Shihao and Guo, Chuan and Zuo, Xinxin and Wang, Sen and Xiaoqin, Hu and Chen, Shoushun and Gong, Minglun and Cheng, Li},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
year={2021}
}