PETS 2014
Seoul, Korea - August 26 2014

In Conjunction with AVSS 2014

ARENA Dataset

Overview


The datasets are multisensor sequences containing  different activities around a parked vehicle in a parking lot.

Note: the dataset was recorded within the framework of the EU project ARENA, whose acronym stands for 'Architecture for the REcognition of threats to mobile assets using Networks of multiple Affordable sensors'.

In this workshop, 22 acted scenarios are provided of abnormal behaviour around the parked vehicle. Although CCTV cameras allow for the surveillance of the area around the vehicle, the main objective is to understand the different behaviours and detect potential threats from four visual (RGB) cameras mounted on the vehicle itself. It is indeed the on-board camera videos that are available for download and analysis. The recorded scenarios are divided into three different threat categories.

Recording characteristics



Physical localisation

The recordings were carried out at the University of Reading, more precisely at the crossing path and car park in front of the School of Systems Engineering. The direct link to the Google maps is as follows: Google Maps .

The CCTV Environmental cameras

Environmental cameras are installed at the locations shown below to cover an approximate area of 100m x 30m. These cameras were active during the scenario recordings. Note that these cameras are not to be processed - rather, they allows authors to obtain a global view of the employed space. Summary low-resolution videos, which include all on-board cameras, and either  enviromental camera 1 or  3,  are available for download.

CCTV cameras at site


Camera installation points are shown in the picture above. There are four cameras available, which specification is as follows:


view Model Resolution frame rate Comments
001 Axis 223M 768x576 ~7 Progressive scan
002 Axis 223M 768x576 ~7 Progressive scan
003 PTZ Axis 233D 768x576 ~7 Progressive scan
004 PTZ Axis 233D 768x576 ~7 Progressive scan


Sample frames from the environmental cameras are shown below:


Sample frames from CCTV cameras 



The On-board cameras

The on-board camera configuration during the recordings is shown below. Four non-overlapping visual cameras are mounted at each corner of the vehicle. They are represented with light blue colour.

on-board cameras disposition


The characteristics for the four on-board cameras are as follows:


Scenario characteristics


Abnormal behaviours

There is a large number of abnormalities recorded in the dataset. However, only some of them are annotated. Behaviours that can be clear indications of abnormalities are:

Threats

Behaviours that are clearly criminal behaviour:


Scenario categories


 

Recording scenarios have been divided into the following categories:


'Something is wrong' scenarios


Sequence ID
Description
Difficulty
Difficulty description
Scene classification
03_06
1 person stands near truck
1
1 person; scene quite empty
Something is wrong
06_01
1 security guard inspecting; walking around truck
1
1 person; scene quite empty
Something is wrong
08_02
Driver falls; someone comes to help
2
2 main persons; scene quiet empty
Something is wrong
03_05
1 person stands near truck someone comes to ask direction
2
2 persons; scene quite empty
Something is wrong
08_03
Driver falls; two persons come to help
3
3 main persons; scene quiet empty
Something is wrong
06_04
Two security guards walk around the truck
3
2 persons; multicamera scenario; scene more frequented
Something is wrong

(clicking on the sequence ID gives a summary video with all recording cameras of the acted scenario.)


'Potentially criminal' scenarios

Sequence ID
Description
Difficulty
Difficulty description
Scene classification
10_04
1 person walks around truck and attempts to open truck
1
1 person; multicamera scenario; scene not too frequented
Potentially criminal
10_05
1 person walks around truck and attempts to open truck
1
1 person; multicamera scenario; scene quite empty
Potentially criminal
11_04
Driver falls pushed by someone
2
2 main persons; 2 more people in pedestrian crossing; Does not look that accidental
Potentially criminal
10_03
1 person walks around truck and attempts to open truck
2
1 person; multicamera scenario; scene somehow frequented
Potentially criminal
11_05
Driver falls pushed by someone
3
2 main persons; Kneeling and falling
Potentially criminal
11_03
Driver falls pushed by someone
3
2 main persons; multicamera scenario
Potentially criminal

(clicking on the sequence ID gives a summary video with all recording cameras of the acted scenario.)
 

'Criminal' scenarios

Sequence ID
Description
Difficulty
Difficulty description
Scene classification
14_01
1 person walking around truck and steals something
1
1 person; scene quite empty; multicamera
Criminal behaviour
14_07
1 Person stealing; loitering; walking around truck
1
1 person; scene quite empty
Criminal behaviour
14_03
1 persons steals something
2
1 person loitering far from truck
Criminal behaviour
14_05
1 person loitering; 1 person stealing
2
2 persons; multicamera
Criminal behaviour
14_06
Group of 3 come to steal; two of them loitering
3
3 persons; multicamera
Criminal behaviour
15_02
Driver hit by two attackers and brought to ground. Attackers take possession of truck
3
3 persons; ; multicamera; scene quite fast
Criminal behaviour
15_05
Driver trapped by asking directions. Driver hit by two attackers and brought to ground. Attackers take possession of truck
3
3 persons; single scene; scene quite empty
Criminal behaviour.

(clicking on the sequence ID gives a summary video with all recording cameras of the acted scenario.) 

Extra 'Criminal' scenarios


Sequence ID
Description
Difficulty
Difficulty description
Scene classification
15_06
Driver involved in a fight with someone. Two more people come to hit driver and brought to ground
3*
4 people fight; complicated scene
Criminal behaviour
22_01
Driver hit and stolen from someone from a car
3*
2 persons; complicated scene
Criminal behaviour
23_01
Driver attacked from someone from a car and brought to floor
3*
2 persons + 1 car; very complicated scene
Criminal behaviour

(clicking on the sequence ID gives a summary video with all recording cameras of the acted scenario.)



Download Information

To download the video sequences (and corresponding tracking results for authors not working on low-level video analysis), please fill out the following User Agreement. You will receive a link to download the dataset after the User Agreement has been registered.

Legal note: The video sequences are copyright University of Reading and permission is hereby granted for free download for the purposes of the PETS 2014 workshop and academic and industrial research. Where the data is disseminated (e.g. publications, presentations) the source should be acknowledged.