无标题文档

Dataset

 

1.HUST American Sign Language (HUST-ASL) dataset [Download]

We use Kinect sensor to collect a new hand gesture dataset including color and depth images with the size of 640*480, named as HUST American Sign Language dataset(HUST-ASL) dataset. Our dataset contains 34 static gestures (10 digits from 0 to 9 and 24 English letters from a to z without dynamic j and z), from 10 people(non-native to sign language) with 16 different poses, which are collected with different degree of deflection in different direction. Therefore, our dataset amounts to a total of 5440 cases.

It should be noted that this dataset is a challenging real-life dataset, which is collected in changing light conditions and cluttered background.The subjects in our dataset are asked to make the sign facing the Kinect device, keep the hand shape fixed, and then move their hand around by rotating clockwise/anticlockwise, or revolving around the wrist, or whirling within a certain degree.

Note:
The folder "Color" contains 10 subfolders ("P1",..., "P10"), which represent ten different people performing gestures. Each subfolders like "P1" contains 34 folders, which represent 34 different gestures performed by one people(10 digits from 0 to 9 and 24 English letters from a to z without dynamic j and z). These 34 folders contain respectively color maps of 16 different poses performed by one people for one gesture.
Just as the folder "Color", the folder "Depth" also contains 10 subfolders ("P1",..., "P10"), which contain 34 folders respectively. These 34 folders contain respectively depth maps of 16 different poses performed by one people for one gesture.

 

Copyright © 2010-2012 Media and Communication Lab

All Rights Reserved

Huazhong University of Science and Technology,Wuhan,P.R.China