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: [email protected]

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}
}