Selection of Visual Descriptors for the Purpose of Multi-camera Object Re-identification - Feature Selection for Data and Pattern Recognition - page 281

Fig. 12.12 Sample images
of persons from HUMAN
dataset

Fig. 12.13 Sample images of two vehicles from the VEHICLE dataset moving in opposite direction
in the fields of view of all cameras
Table 12.4 Characteristics of datasets used in the experiments

Dataset Number Number Number Number of Average Average number Timespan

of of of object/ number of of images per of data

objects cameras images camera images object/camera acquisition

pairs per object pair (h)

HUMAN 7 7 265 48 37.9 5.5 1

VEHICLE 223 8 12,994 1,143 58.3 11.4 9

(descriptor
evaluation
experiments)
VEHICLE 15 8 922 85 61.5 10.9 1

(object
re-identification
experiments)

12.7.2 Descriptor Evaluation

Both datasets have been divided into several subsets based on objects in order to
make analysis results independent of the selection of the specific objects in the
dataset. HUMAN set contains few objects, therefore subsets have been created as all
combinations of five objects out of seven. This gives the total number of 21 subsets.
The VEHICLE set contains an adequate number of objects, therefore the criterion
for dividing is based on the time the object appears for the first time in any camera;
time window length of 1h has been chosen and the window moves with the step of