IntelXVision iSentryStandard-2
IntelXVision iSentryStandard-2
STANDARD
DATASHEET
RELEASE 2023 R2
iSentry
How it Works
Typically, up to 80% of alerts can be handled by the iSentry platform without the need for human intervention. The
inherent risk of automation is largely mitigated by the underlying engine design where rules are applied only when
their results are certain. Key to the iSentry philosophy is that any alert that fails automation will be placed in front of an
operator for further investigation and decision. Together with the rules engine this drastically reduces the number of
false positive alarms by as much as 50 - 70%.
1 2
DECODING AND ANALYSIS DATA ENRICHMENT
iSentry efficiently decodes multiple video streams in A certain number of detected frames, extracted from
real time which are then analysed by one or more of the alert video, are analysed by a GPU-based Deep
the analytics based on Artificial Intelligence, Learning server. This process maximises processing
resulting in an “alert”, which will then start the efficiency and gives the system a much greater
iSentry process. understanding of the alert.
3 4
AUTOMATION LIVE ALERTS
(THE HUMAN COMPONENT)
With the greater understanding gained from step 2, Alerts that are not automatically elevated to alarm in
the system will automatically dismiss many alerts. step 3 are presented to the operator as a list of
current alerts each containing classified images and
a +/- 5-10 second video clip. The operator then
decides whether an alert is important (escalating to
alarm) or not, thus eliminating most false positives.
5
ALARMS AND ESCALATIONS
(TAKING ACTION)
Once an alarm is generated either automatically by
step 3 or by the human in step 4 the iSentry process is
complete. All data associated with the alarm
including video, classifications, metadata, and
operator input, is included with the alarm to be
processed.
Check the
Accuracy and tolerance in real life Configuration &
4 Pixel based Deep Learning environment taking into account difficult Best Practices
camera angles documents
Performance warranted
only on compliance with
Classifying humans using heat signature configuration best practice
24 Human Classification - Thermal Camera from the camera and specified thresholds
for FoV, Object size,
resolution and angle .
Performance warranted
Capable of detecting different sizes of only on compliance with
configuration best practice
25 Fire & Smoke Classification and Detection fires / smoke enhancing early warning and specified thresholds
systems for FoV, Object size,
resolution and angle
Only effective at
Calculating the estimated speed of
26 Object Speed Estimation short range with
classified moving objects
high resolution
Performance warranted
Alert the presence of leaks or spills in only on compliance with
27 Leaks or Spills Detection different environments enhancing safety configuration best practice
and specified thresholds
and environmental monitoring for FoV, Object size,
resolution and angle
Performance warranted
Gender estimation based on appearance only on compliance with
configuration best practice
28 Gender Assessment for retail analytics and demographic
and specified thresholds
analysis for FoV, Object size,
resolution and angle
Only warranted on
compliant camera
36 Group and Crowd Detection Identifying large groups of individuals
positioning, distance
and resolution.
Only warranted on
Identifying prolonged presence of people compliant camera
38 Programmable Loitering Detection
in specific areas positioning, distance
and resolution.
Only warranted on
compliant camera
Identifying adherence to safety gear positioning, distance and
39 PPE Compliance and Reporting protocols
resolution, including but not
limited to gloves, eyewear,
boots or shoes, respiratory
protection devices
Only warranted on
Detection of various types of weapons compliant camera
40 Weapons Detection positioning, distance
including sword, handgun, rifle, shotgun
and resolution.
Identify the presence of banknotes or Only warranted on
currency helping to identify transactions compliant camera
41 Cash Detection involving cash and monitor cash positioning, distance
handling process to enhance security and resolution.
and operational purposes
Identify multiple Wheelchair models
ensuring the wellbeing of individuals
using them in public spaces,
42 Wheelchair Detection
transportation, Malls and Healthcare
facilities by alerting on assistance or aid
in monitoring accessibility
Rules Engine
Rules Engine
Unlimited Boolean logic engine autonomously Unlimited,
decides on alerts for video surveillance using
customised rules, to
Rules and Boolean processing with factors like object types, time, size, and accuracy. It
be applied per scene
46 handles up to 80% of alerts without human input,
unlimited configuration relying on secure logical mechanisms. Alerts failing region.
automated tests are reviewed by human operators
for further action, maintaining a balance between
automation and human oversight.
Other Functions
with appropriate
50 Color Classification Identifying specific colors
lighting conditions