ConductVision Enterprise
ConductVision tracks each person in view and reads their pose, so station, venue and campus teams can time waits and count behaviors by zone and hour.
21 detectionsConductVision detections on this photo. Photo: Asurnipal, CC BY-SA 4.0, cropped.
Behavioral Analytics goes beyond simple object detection to analyze how people and crowds behave in a scene. Built on ConductScience's foundations in behavioral science research, the system can classify complex events such as loitering, aggressive physical interactions, crowd surges, abandoned objects, and unusual movement patterns. These are inherently more challenging than object detection, the system addresses the real-world difficulty of interpreting human behavior from video with transparent confidence scoring on every detection. It is designed as a decision-support tool that alerts trained operators to situations requiring human judgment.
Outcomes
Wait times at counters and gates come from complaints or from a staff member timing a few people.
Extra staff arrive after a queue has built, and the record shows only that people complained.
On platforms and in waiting areas, people stand when the seats run out.
Seating and shelter plans rest on guesses about how many people stand and for how long.
Monthly safety reviews count incidents from staff reports, and reporting varies with who is on shift.
A trend in the review can reflect staffing changes as much as changes on the ground.
Night-shift events reach the day team as a line in a logbook.
Without the footage, the next shift cannot tell a one-off from a pattern.
What the model sees
ConductVision detections on this photo. Photo: Watty62, CC BY-SA 4.0, cropped.
ConductVision detections on this photo. Photo: Ian-Art.photography, CC BY-SA 4.0, cropped.
Endpoints
Each endpoint is defined before the pilot starts, with what is counted, its unit and how the results are broken down.
| Endpoint | How it is counted | Unit | Broken down by |
|---|---|---|---|
| PrimaryBehavior event rate | Events of each behavior the team switches on, such as a long stay in a keep-clear zone, per 1,000 people tracked. | Unit:events per 1,000 people | By:Behavior, zone, hour |
| SecondaryQueue wait time | Minutes from a tracked person's first frame in a queue zone to their last, reported as the median per queue. | Unit:minutes (median) | By:Queue, hour |
| Standing share | Share of people in a waiting area whose pose reads as standing, from sampled frames. | Unit:% | By:Area, hour |
| Long stays | Tracked people who stay in a zone past the time limit set for it, counted once per stay. | Unit:stays per day | By:Zone, hour |
Each pilot measures ConductVision on your own footage against a sample checked by hand, before anyone relies on its numbers.
Behavior alert precision
Share of behavior alerts a reviewer confirms in a hand-checked sample of alert clips.
% · by behavior, camera
Pose agreement
Share of sampled people whose standing, seated or lying label matches a reviewer's label.
% · by camera, pose
Wait time error
Difference between the ConductVision wait time and a reviewer's timing of the same people from the same footage.
seconds (median absolute difference) · by queue
How it works
Connect
Behavioral Analytics works with your existing surveillance cameras. The system requires only standard video feeds, no specialized cameras or sensors. Overlapping camera views improve accuracy for complex behaviors.
Analyze
Unlike single-frame detection, behavioral analysis processes sequences of frames to understand actions and interactions. The system tracks individuals across time, analyzing pose, velocity, proximity, and group dynamics to classify events.
Act
When behavior crosses defined thresholds, a person loitering beyond a time limit, a physical altercation, or a sudden crowd movement, the system alerts security operators with video context. All classifications include confidence scores.
What it detects
Person
Each person in view, tracked with one ID from entering to leaving
Body pose
Keypoints for the head, shoulders, hips, knees and feet of each tracked person
Posture
Standing, seated, bending or lying, read from the pose
Time in zone
First and last frame for each person inside a drawn zone
Speed and heading
Walking pace and direction from each track, for running and counter-flow rules
Features
Use cases
Metro and rail stations need to detect loitering, aggressive behavior, and abandoned bags across dozens of cameras. Behavioral analytics can prioritize feeds requiring immediate attention, reducing operator fatigue.
transportationsmart city
Shoplifting often involves predictable behavioral patterns, loitering near high-value merchandise, concealment gestures, and unusual exit behavior. Anomaly detection can flag suspicious activity for loss prevention staff.
Large venues need to detect crowd surges, bottlenecks, and altercations in real time. Crowd dynamics monitoring can enable safety teams to intervene before dangerous conditions develop.
eventshospitality
Emergency departments and behavioral health units experience elevated rates of aggressive incidents. Early detection of escalating interactions can give security staff time to respond and de-escalate.
City surveillance networks monitor public spaces for safety. Behavioral analytics can transform passive monitoring into proactive alerting, detecting incidents that would be missed in banks of video feeds.
smart citygovernment
Universities and corporate campuses are less populated at night but still require security coverage. Loitering, boundary crossing, and unusual movement detection can supplement reduced overnight guard patrols.
educationcorporate
Parking garages have complex sightlines and multiple levels. Behavioral detection can identify loitering, aggressive encounters, and vehicles circling suspiciously, alerting security to investigate.
Technical overview
Accuracy and speed are measured on your footage and cameras during the pilot, and reported with the conditions they were measured under.
Industries
Counts people, vehicles and bikes at intersections, stops and public spaces.
Tracks forklifts, pallets, trucks and people from dock door to rack.
Tracks PPE, people and heavy equipment across the jobsite.
Counts shoppers, queues and carts from the store's existing cameras.
Measures flow, waits and occupancy in corridors, lobbies and entrances.
Watches crews, trucks and equipment at plants, well pads and tank farms.
Related
Counts and tracks people in each camera view, by zone and hour.
Screening support at entrances and checkpoints, with every alert reviewed by staff.
Finds people, vehicles and animals in thermal camera frames, day or night.
Pilot
A pilot takes one prioritized outcome, runs on your cameras, reports the endpoints above and ends with a results review and a decision.
-queue-01ASD.jpg){kind=link}
