/DeepModel

Code repository for Model-based Deep Hand Pose Estimation

Primary LanguageC++GNU General Public License v3.0GPL-3.0

Model-based Deep Hand Pose Estimation

This repository is the released code of our IJCAI 2016 paper for estimating hand pose from depth image.

Contact: zhouxy13@fudan.edu.cn

Requirements

  • Caffe
  • Python with opencv

Installation

  • Download caffe
  • Copy ./libs/include to caffe_root/include and ./libs/src to caffe_root/src
  • Compile caffe
  • Copy path.config.example to path.config and set the pycaffe path

Test

  • Run demo.py in ./testing
  • Our prediction on NYU dataset here
  • Our prediction on ICVL dataset here

Train

  • download NYU dataset
  • set NYU_path in path.config
  • Run GetH5DataNYU.py in ./training
  • Train with solver.prototxt

Citation

Please cite DeepModel in your publication if it helps your research:

@inproceedings{zhou2016model,
    author = {Xingyi Zhou and Qingfu Wan and Wei Zhang and Xiangyang Xue and Yichen Wei},
    booktitle = {IJCAI},
    title = {Model-based Deep Hand Pose Estimation},
    year = {2016}
}