The
NUS hand posture datasets I & II
Data
Set I
The NUS hand
posture dataset I consists 10 classes of postures, 24 sample images per class,
which are captured by varying the position and size of the hand within the
image frame. Both greyscale and color images are available (160×120 pixels).
The hand postures are selected in such a way that the inter class variation in
the appearance of the postures is less, which makes the recognition task
challenging.
This dataset is
used to test the recognition accuracy of the algorithm reported in the article,
“Pramod Kumar P, Prahlad Vadakkepat, and Loh Ai Poh, Hand
posture and face recognition using a Fuzzy-Rough Approach”, International Journal of Humanoid Robotics, vol.7, no.3,
pp.331-356, September, 2010. The dataset can be used for academic research
purposes free of cost, by citing the article.
PS: The
background of the images in this dataset is uniform. Another hand posture dataset containing
images with complex background is available in the NUS Hand Posture Dataset-II.
Download NUS Hand Posture Dataset I (Zipped file)
Data
Set II
This is a 10
class hand posture dataset. The postures are shot in and around National
University of Singapore (NUS), against complex natural backgrounds, with
various hand shapes and sizes. The postures are performed by 40 subjects, with
different ethnicities, against different complex backgrounds. The subjects
include both males and females in the age range of 22 to 56 years. The subjects
are asked to show the 10 hand postures, 5 times each. They are asked to loosen
the hand muscle after each shot, in order to incorporate the natural variations
in the postures.
The dataset also
consist of a set of background images which does not contain any of the hand
postures.
Notes:
1.
The folder 'Hand Postures' contains the 10 class hand posture
images (160x120 size). Classes 1 to 10 are named with starting letters 'a' to
'j' respectively. A total of 2000 hand
posture images are present in this folder.
2.
The folder 'Hand Postures with human noise' contains the 10 class
hand posture images (320x240 size) with human noises (face of the posturer,
human in the background etc). Classes 1 to 10 are
named with starting letters 'a_HN' to 'j_HN' respectively.
A total of 750 hand posture images are present in this folder.
3.
The folder 'Backgrounds' contains the background images. A total
of 2000 background images are available.
4.
All the images are RGB and in .jpeg format.
5.
The hand posture detection and recognition results using this
dataset are reported in the paper: Pramod Kumar Pisharady,
Prahlad Vadakkepat, Ai Poh Loh, "Attention Based Detection and Recognition of Hand Postures Against Complex Backgrounds", International
Journal of Computer Vision, vol.101, no.3, pp.403-419, February, 2013. The
dataset can be used for academic research purposes free of cost, by citing the above
article.
Download NUS Hand Posture Dataset II
(Zipped file)