Organizer
Paul Hosmer, Home Office Scientific
Development Branch (UK)
Description
Intelligent detection systems that provide automatic event
detection using CCTV imagery have huge potential to
increase the effectiveness of CCTV systems used for
security and surveillance. Imagery is made available for
researcher in this area via the UK government's i-LIDS
standard.
This special session aims to attract papers that present
state of the art methodologies for event detection. It
intends to report on the accuracy, robustness and
complexity on a set of clips extracted from the i-LIDS
dataset. We encourage the submission of papers in
the following topics:
-
Abandoned baggage detection
- Parked vehicle detection
The Home Office Scientific Development Branch (HOSDB) are
offering either an i-LIDS Abandoned Baggage or
Parked Vehicle training dataset as a prize to the group or
individual whose algorithm offers the best performance on
unseen clips from the i-LIDS library. These clips will be
from the same cameras as those distributed for the AVSS
2007 conference and participants will be required to
detect the events described by the i-LIDS scenario
definitions.
The demonstration on the unseen clips will take place
during the AVSS 2007 conference. The winner will be
determined based on accurate detection of the events and
the number of false alarms generated during the sequence.
The
description of the tasks can be found
here.
The
dataset of the i-LIDS bag and vehicle detection challenge
is available
here
The dataset contains three clips
from the Abandoned Baggage scenario and four clips from
the Parked Vehicle scenario.
To participate in the challenge, authors can choose to
process the whole dataset or part of the dataset. The
event detection results (StartTime, AlarmDescription, and
AlarmDuration) for each processed clip should be reported
in the submitted paper in the form of a table. Note that
no XML submission is required at the paper submission
stage. Participants should declare in their paper if they
have access to the full i-Lids dataset.
The challenge will take place during the AVSS 2007
conference in September, when unseen data from the i-Lids
dataset will be used to test and rank the algorithms.
More information on i-LIDS can be found in the CCTV and
Imaging Technology area of the
HOSDB website
Calibration images
Calibration images for the unseen challenge: (i)
parked vehicle (jpg, 345 KB) (ii)
abandoned
baggage
(jpg, 300KB)
Briefing document
The AVSS 2007 i-LIDS challenge brief document can be found
here
Contact
Contact email:
challenge@avss2007.org
|