Start with the source.
Agree on the question, authorized footage, relevant zones, and operating conditions. Camera compatibility and local processing requirements need assessment for each site.
SENTINEL VISION / ACTIVE DEVELOPMENT
From visible activity
to useful understanding.
A modular visual-intelligence platform being developed to turn authorized camera footage into observations people can inspect, question, and use.
THE PROPOSED WORKFLOW
A shared foundation.
A specific operational question.
Develop useful signals.
Evaluate candidate counts, activity patterns, or events against reviewed examples. Keep uncertainty visible and connect observations to evidence where feasible.
Test against reality.
Compare outputs with human review and an agreed baseline. Decide whether the results answer the original question before expanding scope.
APPLICATIONS UNDER EXPLORATION
Concrete questions.
Measurable pilot work.
01 / RETAIL & VENUES
Understand how a space is used.
When do entrances become busy, and where do visitors spend time?
Proposed scope
Explore entry and exit counts, zone occupancy, and dwell patterns from an authorized view of a defined area.
What needs validation
Compare observations with manual counts. Test crowding, occlusion, camera placement, and changing light before relying on trends.
02 / DEALERSHIPS & PROPERTY
Put lot activity in context.
Which areas attract activity, and how does traffic change through the day?
Proposed scope
Explore vehicle and pedestrian movement across agreed lot zones, with time-based summaries and reviewable footage references.
What needs validation
Separate passing traffic from on-site activity. Check night conditions and overlapping vehicles; movement alone does not establish purchase intent.
03 / FACILITIES & OPERATIONS
Review the events that matter.
What activity occurs around an access point or a restricted operational zone?
Proposed scope
Explore zone-based event candidates and supporting visual context for staff review, using site-defined rules and authorized footage.
What needs validation
Measure missed events and false positives with operators. Any alerts or workflow integrations require separate scoping and validation; this is not a certified safety system.
04 / WILDLIFE & LAND
Make observation easier to review.
When are animals active, and which observations deserve a closer look?
Proposed scope
Explore animal-activity timelines and species candidates from fixed cameras or recorded clips, with evidence retained for human review.
What needs validation
Check species candidates against expert labels. Weather, vegetation, distance, and infrared imagery can change results; candidates are not confirmed identifications.
05ENGINEERING / EVIDENCE
BEHIND THE CONCEPT
Built against real observations.
Measured before promised.
Sentinel's current wildlife research is a proving ground for the underlying video-analysis infrastructure. We are building the tools to connect proposed detections to reviewable evidence, not assuming every model proposal is correct.
REVIEW REGION
+
CONCEPTUAL REPRESENTATION — NO CAMERA FOOTAGEFind candidate activity
Research workflows use recorded and live camera inputs to investigate motion, object proposals and temporal consistency. False alerts and missed detections remain active engineering challenges.
EVENT EVIDENCE
Candidate proposalNeeds review
Object localizationUnverified
Species hypothesisUnknown
Human feedbackRequired
ILLUSTRATIVE WORKFLOW, NOT LIVE MEASUREMENTSPreserve uncertainty
The system records hypotheses separately from human conclusions. A detected box does not automatically establish a verified animal, species, visit or unique individual.
VALIDATION CHECKLIST
01Representative footage○
02Ground-truth review○
03False-positive analysis○
04Repeatable benchmark○
REQUIRED BEFORE PRODUCT PERFORMANCE CLAIMSProve the use case
For people counting, facility events or wildlife observation, success must be measured with agreed samples and human-reviewed results before commercial claims are made.
Have a camera-based question that matters to your operation? We are looking for focused evaluation opportunities.
Explore an early partnershipREAL-WORLD BOUNDARIES
The environment is part of the problem.
Footage, compute, and reliability
Resolution, viewing angle, lighting, bandwidth, and available hardware affect feasibility. Local-first processing is a design direction; deployment architecture, retention, uptime expectations, and integrations must be agreed for a specific pilot.
Permission and human judgment
Partners must have authority to use the footage. Access, privacy, retention, and review responsibilities should be defined before any data is shared. Observations need human interpretation and do not establish identity, intent, or safety.