|-----------------------------------------------------------------------------| |------------VAP Trimodal People Segmentation Database Info ------------------| |-----------------------------------------------------------------------------| | | | The dataset comprises of 3 scenes, named as follows: | | 1: Meeting room full depth | | 2: Meeting room constrained depth | | 3: Canteen | | | | The dataset contains activities of up to three people captured by a | | thermal, depth, and colour sensor. The activity is constrained within the | | depth range of the Kinect for Windows V1 sensor in "normal mode". | | | | The cameras used the capturing were | | Axis Q1922 | | Kinect for Windows V1 | | | |-----------------------------------------------------------------------------| | | | A registration dataset for mapping points between the RGB, depth, and | | thermal modalities are included in the C++ files registrator.h, | | registrator.cpp containing functions for the mapping of points and con- | | tours. | | Required libraries are OpenCV version 2.4 or higher and the dirent.h | | framework. For compilation on Windows, dirent.h should be downloaded | | separately. | | | | The registration of each scene is dependent on variables stored in | | calibVars.yml which is unique for each scene. | | The use of the registration algorithm is described in registrator.h | | and demoed in the registrator.cpp main function. | | | |-----------------------------------------------------------------------------| | | | The thermal images are stored as a 32bit bitmap of resolution 680x480 px. | | The thermal images suffer from radial lens distortion which may be | | rectified by the use of the registrator.h/cpp files in combination with | | the corresponding calibVars.yml file. | |-----------------------------------------------------------------------------| | | | The color images are stored as a 32bit bitmap of resolution 680x480 px. | | | |-----------------------------------------------------------------------------| | | | The depth images are stored as 16-bit png images where each pixel in | | the image represents the depth of that corresponding pixel in mm. | | Values for the valid depth are in range from 1000 - 3300. | | Other values are constants specified by the Kinect. | | -1 - Undefined distance | | 0 - Too near | | 4095 - Too far | | | | | |-----------------------------------------------------------------------------| |-----------------------------------------------------------------------------|