Content
This dataset was recorded due to the lack of
public parking car datasets. The sequences were
recorded in a real environment of the Pittsburgh International Airport
parking, in order to work with an environment as realistic as possible.
Each of the frames has a resolution of 1280x960 pixels, recorded using Panasonic WV-SW155 cameras. The dataset consists of two main image sets,
a training set used to generate the detector models, and a test set
used for the experimental evaluation. The training set consist of a
longer set of images, and the test set consist of a long (named 'All')
and a short (named 'Synchronized') version of the images, with 1000 and
100 frames, respectively. The short versions (Synchronized sets) are
contained in the long version and have the frames synchronized between
the two cameras, to be able to evaluate experiments combining the
information of both cameras. In addition to generating the images, the
vehicles of all images have been manually annotated. The training
images have been annotated for its use in the generation of the parked
vehicle model, and the test images for the evaluation of the system. In
the case of the synchronized multicamera set, the vehicle occupancy
matrix has been manually generated. |