Fast and Robust Dynamic Hand Gesture Recognition via Key Frames Extraction and Feature Fusion.
Hao Tang1, Hong Liu2, Wei Xiao3 and Nicu Sebe1.
1University of Trento, Italy, 2Peking University, China, 3Lingxi Artificial Intelligence Co., Ltd, China.
In Neurocomputing 2019.
The repository offers the official implementation of our paper in MATLAB.
Copyright (C) 2019 University of Trento, Italy.
All rights reserved. Licensed under the CC BY-NC-SA 4.0 (Attribution-NonCommercial-ShareAlike 4.0 International)
The code is released for academic research use only. For commercial use, please contact [email protected].
Clone this repo.
git clone https://github.com/Ha0Tang/HandGestureRecognition
cd HandGestureRecognition/
This code requires MATLAB. Please install it.
For Cambridge Hand Gesture or Northwestern Hand Gesture, the datasets must be downloaded beforehand. Please download them on the respective webpages. Please cite their papers if you use the data.
Preparing Cambridge Hand Gesture Dataset. The dataset can be downloaded here. You can also download this dataset use the following script:
bash ./datasets/download_handgesture_dataset.sh Cambridge_Hand_Gesture
Preparing Northwestern Hand Gesture Dataset. The dataset is proposed in this paper. You can download this dataset use the following script:
bash ./datasets/download_handgesture_dataset.sh Northwestern_Hand_Gesture
Preparing HandGesture Dataset. This dataset consists of 132 video sequences of 640 by 360 resolution, each of which recorded from a different subject (7 males and 4 females) with 12 different gestures (“0”-“9”, “NO” and “OK”). Download this dataset use the following script:
bash ./datasets/download_handgesture_dataset.sh HandGesture
Preparing Action3D Dataset. This dataset consists of 1620 image sequences of 6 hand gesture classes (box, high wave, horizontal wave, curl, circle and hand up), which are defined by 2 different hands (right and left hand) and 5 situations (sit, stand, with a pillow, with a laptop and with a person). Each class contains 270 image sequences (5 different situations × 2 different hands × 3 times × 9 subjects). Each sequence was recorded in front of a fixed camera having roughly isolated gestures in space and time. All video sequences were uniformly resized into 320 × 240 in our method. Download this dataset use the following script:
bash ./datasets/download_handgesture_dataset.sh Action3D
New models can be trained with the following commands.
-
Prepare your own dataset like in this folder.
-
Extract key frame:
matlab -nodesktop -nosplash -r "key_frames_extraction"
Key frames will be extrated in the folder ./datasets/sample_keyframe
.
- Go this folder for further processes.
- Clustering by Fast Search-and-Find of Density Peaks
- Gender Classification using Pyramid Segmentation for Unconstrained Back-Facing Video Sequences
If you use this code for your research, please cite our papers.
@article{tang2019fast,
title={Fast and Robust Dynamic Hand Gesture Recognition via Key Frames Extraction and Feature Fusion},
author={Tang, Hao and Liu, Hong and Xiao, Wei and Sebe, Nicu},
journal={Neurocomputing},
volume={331},
pages={424--433},
year={2019},
publisher={Elsevier}
}
This work is partially supported by National Natural Science Foundation of China (NSFC, U1613209), Shen- zhen Key Laboratory for Intelligent Multimedia and Virtual Reality (ZDSYS201703031405467), Scientific Research Project of Shenzhen City (JCYJ20170306164738129).
If you have any questions/comments/bug reports, feel free to open a github issue or pull a request or e-mail to the author Hao Tang ([email protected]).