***************************************************
***						***
***	ColorTip Dataset			***
***						***
***	23 - Oct - 2012				***
***						***
***	Universitat Politècnica de Catalunya	***
***						***
***	contact: xavier.suau@upc.edu		***
***						***
***************************************************

ColorTip is a Computer Vision Dataset for hand gesture recognition and fingertip classification using Kinect.

=============
DATA PROVIDED
=============

The dataset consists of 7 subjects performing 2 sequences each: set_A and set_B (more challenging gestures), at a distance of approximately 100cm far from the camera.

For each sequence, depth and color images are provided, as well as groundtruth annotations for the hand gesture, hand position and fingertip positions and labels (see GROUNDTRUTH ANNOTATIONS below).

The dataset is organized as follows:

capture/subject-01/d100/set_A/image/seq_%06d.png	--> RGB images
capture/subject-01/d100/set_A/depth/seq_%06d.png	--> Depth images (uint16 format)
capture/subject-01/d100/set_A/gt/			--> groundtruth files (annotation.txt and gestures.txt)
 
capture/subject-01/d100/set_B/image/seq_%06d.png	--> RGB images
capture/subject-01/d100/set_B/depth/seq_%06d.png	--> Depth images (uint16 format)
capture/subject-01/d100/set_B/gt/			--> groundtruth files (annotation.txt and gestures.txt)

.
. (idem with subjects 02 to 07)
.


=======================
GROUNDTRUTH ANNOTATIONS
=======================

annotation.txt
==============

Hand (h) and Fingertip (fk, k=1..5) annotations, in Kinect image coordinates.

format: frame hx hy f1x f1y f2x f2y f3x f3y f4x f4y f5x f5y

a -1000 value stands for a missing annotation.

gestures.txt
============

Hand Gesture annotation. Transitions between gestures are indicated:

format: gesture transition_frame
i.e.

9 0	--> gesture 9 starts at frame 0
1 45	--> gesture 1 starts at frame 45
2 185	--> gesture 2 starts at frame 185
and so on...