Security camera video data: events and look-alikes
The moments security cameras must tell apart: a skipped scan and an honest checkout, a fall and someone sitting down, an intruder and a routine visitor, and the drills, umbrellas and phones that set off false weapon alerts. Incidents are staged by consenting adult actors at private sites with harmless everyday objects, never real weapons, and every clip is labelled as the event or its look-alike, with start and end times.
Specification
- Footage
- Security-camera angles (high, wide, ceiling and doorway) at private shops, homes, offices and car parks, by day and at night
- Events
- Checkout skip-scans and pass-arounds, falls, loitering, forced entry and tailgating through doors
- Look-alikes
- Honest checkouts, sitting, kneeling and bending down, waiting, deliveries and visitors, and drills, umbrellas, phones and tools in hand
- Actors and sites
- Consenting adult actors at private sites, with harmless everyday objects only, never real weapons; anyone else in view blurred
- Annotation
- Event or look-alike, event type, start and end times, boxes for people and objects, camera angle and light
- Quality check
- Every clip is cross-checked by several trained reviewers
Example of the data
What you see, and the file you get.
| skip-scan | event | item passed around the scanner |
|---|---|---|
| honest scan | look-alike | item scanned and bagged |
| cordless drill | look-alike | held in hand, not a weapon |
| video_file | camera | start | end | label | event_type | note |
|---|---|---|---|---|---|---|
| cam_0204.mp4 | checkout_overhead | 12 | 15.4 | event | skip_scan | item passed around the scanner |
| cam_0204.mp4 | checkout_overhead | 31.2 | 33 | look_alike | honest_scan | item scanned and bagged |
| cam_0217.mp4 | hallway_corner | 8.1 | 10.9 | event | fall | person ends up on the floor |
| cam_0217.mp4 | hallway_corner | 21.5 | 24 | look_alike | sit_down | person sits on a low chair |
| cam_0233.mp4 | entrance_wide | 47.3 | 52.1 | look_alike | object_in_hand | cordless drill, not a weapon |
| cam_0241.mp4 | back_door | 3.4 | 9.8 | event | tailgating | second person follows a badge holder in |
Sample rows. Also as JSON or Excel.
Contact us to get the dataset
Tell us what you need. We walk you through the data, its formats and how it is checked on a short call.
What it is used for
Retail loss prevention
Catch skipped scans at self-checkouts without flagging honest shoppers.
Fall detection
Tell a fall from sitting, kneeling or bending down, at home and in care settings.
Fewer false alarms
Teach detectors the drills, umbrellas and phones that set off false weapon alerts.
Video analytics testing
Measure how often a model mistakes everyday moments for incidents.
Questions
Are the incidents real?
They are staged by consenting adult actors at private sites, using harmless everyday objects and never real weapons. The look-alikes are everyday moments, filmed the same way from the same camera angles.
Which events and look-alikes are covered?
Skip-scans and honest checkouts, falls and sitting down, intruders and routine visitors, and everyday objects that weapon detectors mistake for threats. Your own event list can be used.
Can I see samples first?
Yes. Request samples with the form on this page and we will send them with the full specification.
Tell us what your model needs.
Request samples and the full specification. We reply within a couple of business days.
Request samples