Skip to content

Code for paper <AWR: Adaptive Weighting Regression for 3D Hand Pose Estimation>. Accepted by AAAI 2020.

License

Notifications You must be signed in to change notification settings

Elody-07/AWR-Adaptive-Weighting-Regression

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AWR: Adaptive Weighting Regression for 3D Hand Pose Estimation

New:

  • 2021/11/30: We have updated our code and integrate test & train code both in train.py. To validate our model, simply run train.py and you should get the result for hourglass_1.

  • An update version of paper is here. We re-compute the params and FLOPs of each network structure, mainly on Hourglass and update the results accordingly.

This is the official repository for AAAI 2020 paper AWR: Adaptive Weighting Regression for 3D Hand Pose Estimation. by Weiting Huang*, Pengfei Ren*, Jingyu Wang, Qi Qi, Haifeng Sun (* denotes equal contribution)

Codes are implemented with Python 3.7.0 and Pytorch 1.4.0.

Introduction

We propose an adaptive weighting regression (AWR) method to leverage the advantages of both detection-based and regression-based method. Hand joint coordinates are estimated as discrete integration of all pixels in dense representation, guided by adaptive weight maps. This learnable aggregation process introduces both dense and joint supervision that allows end-to-end training and brings adaptability to weight maps, making network more accurate and robust.

Fig 1. Main idea of AWR.

Fig 2. Framework of AWR

Code Setup

We provide result on NYU dataset with Resnet18 (resnet_18_uvd.txt) and inferencing code.

  1. Download NYU dataset and put train and test directory in ./data/nyu. We also provide hand center trained using a separate 2DCNN.
  2. pip install -r requirements.txt
  3. Modify ./config.py according to your setting.
  4. Run code python test.py

New:

  • We update training and data augmentation code for reference.
  • We provide HourGlass network and a pretrained model on HourGlass-1stage. You can set load_model='./results/hourglass_1.pth' to run the code.

Citation

If you find our work useful in your research, please citing:

@inproceedings{awr,
  title={AWR: Adaptive Weighting Regression for 3D Hand Pose Estimation},
  author={Weiting Huang and Pengfei Ren and Jingyu Wang and Qi Qi and Haifeng Sun},
  booktitle={AAAI Conference on Artificial Intelligence (AAAI)},
  year={2020}
}

About

Code for paper <AWR: Adaptive Weighting Regression for 3D Hand Pose Estimation>. Accepted by AAAI 2020.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages