01 / 04
Item hidden in a bag
A shopper puts an item into a bag or backpack instead of a cart or basket.
- Hand moves from the shelf to a bag
- The item does not come back out
- The bag stays with the shopper
HOW IT WORKS
It reads body movement, not faces.
Omicent follows every shopper on every camera, reads hands and posture frame by frame, and flags the moment an item disappears into a bag, a pocket or clothing. Your team gets the clip in 30 to 60 seconds.
The pipeline
Four layers, one real clip.
Two shoppers in a power tools aisle. Step through what Omicent sees, from following each person to the alert on a phone.
What it catches
Named behaviors, not vague “suspicious activity”.
Each alert is a clear yes or no for your team. These are real store clips, faces blurred, with the body points Omicent reads drawn on top.
02 / 04
Item hidden under clothing
An item goes under a shirt or inside a jacket.
- Hand moves under the shirt or jacket
- The item goes out of sight on the body
- Often while facing the shelf
03 / 04
Item tucked into a waistband
A small item goes into a waistband or a pocket.
- Both hands move to the waist
- Small items that fit in a hand
- Standing close to the shelf
04 / 04
Shoppers working together
One shopper hides an item on another, or in the other person’s bag.
- Two shoppers close together
- An item passes between them
- It ends up hidden on the second
It also watches the lead-up.
These signals tell Omicent where to look closer. On their own they are not alerts, so your team is not buzzed for browsing.
Reaching and crouchingIn aisles you mark as high-theft
Lingering at high-value displaysTime at cases and shelves you choose
Many items taken at onceSeveral items pulled off a shelf in one go
A path that skips the checkoutFrom the aisle toward the exit, past the registers
A group that splits upPeople who arrive together and spread out
Leaving with goods, no saleWhen your POS data is connected
Every kind of store
Shoe stores, hardware aisles, beauty counters.
The same body-level signals work across formats, because hiding an item looks the same in any aisle.
Footwear
Items hidden at the waist
At the try-on bench, small items come off the shelf and go to the waist.
Hardware
Item to pocket
In the power tools aisle, a hand moves an item toward a pocket while a partner waits.
Beauty
Open tote at the shelf
A shopper with an open tote works along the skin care shelves.
Pharmacy
Lingering in a high-theft aisle
A long stay at the deodorant shelf tells Omicent where to look closer.
Accessories
Item held close to the body
Small packaged items come off a display and stay close to the body.
Apparel basics
Pack bundled to the body
At a socks display, a pack is unfolded and bundled against the body.
What your team gets
Every alert is a short clip your team can act on.
Store and cameraWhich store and which camera saw it
TimeWhen it happened, to the second
What was seenThe behavior, in one plain sentence
The clipCentered on the shopper, ready to play on a phone
Where alerts go
Store manager phonesLoss prevention teamEmailYour security systemWebhooks and API
Offer customer serviceA staff member walks over and offers help. Often that is enough.
Watch the exitKeep an eye on the door as the shopper leaves.
Log it for follow-upEvery clip is saved, so nothing depends on catching it live.
Your policy, your call. Omicent shows what happened. What staff do with it follows your store’s rules.
2:47●●● ▮
Omicent now Possible theft · Shoppers working together
- Store
- Store 0214 · Downtown
- Camera
- CAM 02 · Try-on bench
- Time
- 2:47:13 PM
"Shopper placed an item from the shelf into a backpack."
WatchEscalateShare clip
Setup
Live in days. No new hardware.
Omicent connects to the cameras and recorder you already run. Most stores need nothing installed at all.
01
Connect your recorder
We connect to your NVR or IP cameras over a secure RTSP link to the Omicent cloud. Works with most recorders, including Hikvision, Axis, Hanwha and Avigilon.
02
Mark the store
Together we mark the checkout, the exits and the aisles with high-theft items, and choose who gets alerts.
03
Alerts go live
Clips start reaching the phones you chose. Every incident is saved for your weekly report.
Camera views
What makes a camera work well.
Most existing cameras work as they are. During setup we check each view and may suggest moving a few, or adding one over a high-theft aisle.
Works well
- Sees the shopper’s hands and the shelf
- Looks along the aisle from above head height
- Covers aisles with high-value or high-theft items
- Covers the exits and the path past the checkout
- Even lighting, no strong glare
Needs a change
- Only sees the tops of heads from straight above
- So wide that a shopper is a few pixels tall
- Shelves block the view of hands
- Backlit by glass doors or windows
Always improving
It gets sharper every week.
Omicent learns from real store video across many retailers, so every new store starts with everything learned so far.
Real store videoModels train on genuine retail footage, not staged lab clips.
Every outcome countsEach confirmed theft and each false alarm sharpens the next decision.
Models updateImprovements roll out to every store automatically, with no work on your side.
Fewer false alertsAlert quality rises over time, and new behaviors are added as patterns change.
Privacy built in
It watches movement, not identity.
No face recognition
Omicent does not identify anyone. It reads how hands and bodies move.
Temporary IDs
Each shopper gets an ID that resets when they leave the camera.
No tracking across visits
Nobody is followed across visits or across stores.
Only flagged clips are kept
Short clips of flagged moments, nothing else, retained on your policy.
Encrypted in transit and at rest
Video and clips are protected with TLS in transit and AES-256 at rest.
You control access
You decide who sees alerts and clips, store by store.
Why Omicent
Built to catch the act, not just record it.
01
Body-level detection, not motion alerts
Omicent reads hands, arms and posture on every shopper, so it knows the difference between browsing and hiding an item. Motion detection can’t.
02
Named behaviors your team can act on
Item into a bag, under clothing, into a waistband, passed to a partner. Every alert says what was seen, so staff know what to look for.
03
Every alert is a clip, not a guess
A few seconds of video centered on the shopper, starting just before the moment, ready to play on a phone.
04
Sharper every week
Models learn from real store video across many retailers, so a new store starts with everything learned so far.
FAQ
The questions retailers ask us first.
Omicent is AI shoplifting detection software for retailers. It connects to a store’s existing cameras, reads shoppers’ body and hand movement, and sends staff a short video clip within 30 to 60 seconds when someone hides merchandise in a bag, under clothing or in a waistband. It does not use facial recognition.
Omicent follows each shopper with a temporary ID, reads 17 body points per frame, and flags concealment, such as a hand moving from the shelf to a bag where the item does not come back out. Each flagged moment becomes a few seconds of video centered on the shopper, sent to the store team with the store, camera and time.
Usually not. Omicent works with the cameras and recorder you already have. During setup we check each view and may suggest moving a few cameras, or adding one over a high-theft aisle.
Not for most stores. Omicent connects to your NVR or IP cameras over a secure RTSP link and runs in the cloud. For large stores and chains we can place a small Omicent box on site, one per 20 to 30 cameras, so full video stays in the store.
Within 30 to 60 seconds of the moment, so staff can act while the shopper is usually still in the store. Every clip is also saved, so nothing depends on catching it live.
No. Omicent looks at movement, not identity. Nobody is identified, and nobody is followed across visits or stores.
Omicent alerts on named behaviors, like an item going into a bag, not on vague “suspicious” movement. Browsing, crouching or lingering on their own never trigger an alert. The models also improve every week from real store video.
Pricing depends on the number of stores and cameras. We’ll walk you through it on a 15-minute demo, along with a pilot plan for your own stores.
Omicent is a Dragonfruit AI company, headquartered in San Jose, California. It is built on Dragonfruit’s video AI, which runs in production at more than 15,000 sites.
Ready when you areSee it catch theft on
See it catch theft on
your own cameras.
Book a 15-minute demo. Then pilot it on the stores where you lose the most.
Book a demo