SCUT-DHGA

This is the official repository of "Dynamic Hand Gesture Authentication Dataset and Benchmark".

Dataset

We introduce a new dataset SCUT-DHGA which is the first large-scale Dynamic-Hand-Gestures authentication dataset. SCUT-DHGA contains 29,160 dynamic-hand-gesture video sequences and more than 1.86 million frames for both color and depth modalities acquired from 193 volunteers. Six kinds of dynamic hand gestures are carefully designed: 1)make a fist starting from thumb, 2)rotate hand while making a fist starting from little finger, 3)catch and then release, 4)four fingers(index, middle, ring, little) touching thumb one by one, 5)bend four fingers(same as 4)) one by one, 6)open a fist starting from thumb.

Now a small part of our dataset containing both depth and color gesture videos from five subjects can be downloaded here. Note that the depth data is 8-bit in this demo version. The 16-bit depth data will be released soon in our complete dataset.

[Update] We have released our complete dataset!

Request

The SCUT-DHGA dataset is publicly available (free of charge) to the research community.
Unfortunately, due to privacy reasons, we cannot provide the database for commercial use.

Those interested in obtaining SCUT-DHGA should download release_agreement, and send by email one signed and scanned copy to [email protected].

While reporting results using the SCUT-DHGA, please cite the following article:
@ARTICLE{9249008, author={Liu, Chang and Yang, Yulin and Liu, Xingyan and Fang, Linpu and Kang, Wenxiong}, journal={IEEE Transactions on Information Forensics and Security}, title={Dynamic-Hand-Gesture Authentication Dataset and Benchmark}, year={2021}, volume={16}, number={}, pages={1550-1562}, keywords={Authentication;Feature extraction;Training;Benchmark testing;Biometrics (access control);Physiology;Dynamic-hand-gesture;authentication;dataset;two-sessions;benchmark}, doi={10.1109/TIFS.2020.3036218}}

How to run

1. Prepare the code and python environment

git clone https://github.com/SCUT-BIP-Lab/SCUT-DHGA.git
cd SCUT-DHGA
pip install -r requirement.txt

2. Pepare the dataset and place it somewhere

3. Just run the code

for training

python main.py \
--training_file $path to the training config file$ \
--testing_file $path to the test config file$ \
--data_root $path to the dataset$ \
--train

for testing

python main.py \
--training_file $path to the training config file$ \
--testing_file $path to the test config file$ \
--data_root $path to the dataset$ \
--test \
--testmodel_name $path to the parameters$