A fashion chain, a QSR chain, a logistics company and a steel plant. Different floors, the same edge device, and in the steel plant a check where being right most of the time is not enough.
Who uses Tarsyer?
Tarsyer runs in retail, restaurant, logistics and manufacturing businesses across India. They include a fashion chain with more than 1,200 stores, a QSR chain with more than 300 restaurants and 5,000 cameras, and a logistics company with an edge device at every loading dock. A steel plant uses Tarsyer to check that each crane hook is seated before a ladle is lifted.
Fashion retail
One score for every store, every morning.
1,200+stores on one system
1,000+store managers using it daily
1device per store
The problem
With more than a thousand stores, head office could not see what happened on each floor between visits: how many people walked in, whether checkout queues were building, whether fire exits were kept clear.
What we put in
A Tarsyer device in each store, reading the cameras and recorder already installed. Footfall is counted at the entrance with staff separated out. Cameras at the tills watch the queues, and others cover the fire exits, the back office and the entrance after hours.
What changed
Every store manager starts the day with the store’s score, the footfall trend against yesterday and last week, and the specific issues that cost points. A nightly review turns each store’s numbers into a short list of things to fix that day. Area managers see their stores ranked, and head office sees the whole chain.
Footfall with staff excludedCheckout queuesFire exitsBack officeAfter-hours intrusionDaily store score
The screen each store manager opens.
Restaurants and QSR
Half the problems turned out to be one problem.
300+restaurants
5,000cameras watched
12standard checks
The problem
A national QSR chain had cameras in every restaurant and nobody with time to watch them. Table cleaning, caps and gloves in the kitchen, and a staffed counter all depended on someone being there to notice.
What we put in
A device per restaurant running the twelve standard checks on the cameras already there. Tables, counters, bins and doors were outlined once on each camera view. Every detection is re-read by a second vision model before the alert goes out on WhatsApp with the photo.
What changed
Once every issue was counted, the pattern was plain. In one week across 37 restaurants, 375 of the 686 confirmed issues were tables left uncleaned, more than half of the total. That told the chain where to spend its effort. Camera health is tracked on every camera too, so a dark or moved camera is fixed before its footage is needed.
Table not cleanedNo staff at the counterCaps and glovesPhone use in the kitchenUnauthorised entryCamera health
Confirmed issues by type across the chain.
Warehousing and logistics
The dock was idle more often than it was busy.
1device per dock
0changes to the dock workflow
0extra staff to run it
The problem
A logistics company wanted to know what was happening at each loading dock without sending supervisors to walk the floor: which docks were working, which were waiting, and whether safety rules were kept at the dock edge.
What we put in
A mobile-phone sized edge device at each dock, reading the existing cameras. Nothing changed in the dock workflow, the network or the camera positions. Each dock is classed every second as active, idle or empty, alongside pallets handled, trucks turned round, and safety checks on shoes, reflective jackets and phones.
What changed
The split of dock time made the case on its own. In the period shown on the dashboard, docks were idle 54% of the time, active 34% and empty 12%. Idle time, when a dock that could be working is not, is the time an operations team can win back.
Dock active, idle, emptyPallets handledTrucks turned roundSafety shoes and jacketsPhone use at the dockLoitering
Dock time, pallets, trucks and safety counts on one screen.
Steel manufacturing
The crane does not lift until the hook is proven safe.
30 tof molten metal in each ladle
2camera views of every hook
1edge device per camera
The problem
In a steel plant, overhead cranes lift ladles holding around thirty tonnes of molten metal. The crane hooks onto pins on the side of the ladle, called trunnions. If a hook is not seated properly, the ladle can tip as it rises. The check has always rested on the operator’s judgement from the cab, high above.
What we put in
Cameras at each ladle pick-up point, placed so every hook is seen from both sides of the trunnion, each with its own edge device. The model finds the hook and the trunnion on every frame and measures how far they overlap. It also recognises which crane is over the ladle from the crane’s shape. A unit in the crane cab shows the operator the result in real time, and every lift is kept on video for review.
How it stays safe
A check guarding a lift like this cannot rely on being right most of the time, so it is built to fail safe. The cab shows a warning from the moment the crane arrives and turns safe only when the overlap passes a threshold set in advance. A misplaced hook shows as unsafe. An uncertain reading therefore holds the lift rather than clearing it, and the operator still acknowledges the safe signal before lifting. A few seconds’ wait is an acceptable cost. Clearing an unsafe hook is not.
Where it is now
The check is live at the plant. Each lift is confirmed in the crane cab before the ladle rises, and every lift is kept on video for review.
Hook seated on the trunnionCrane identified by shapeSafe and unsafe signal in the cabOperator acknowledgementVideo kept for every liftCamera and device health
How the check works. The model measures the overlap between hook and trunnion, and the cab shows the result.
Your business could be the next story.
Start with one site. We install on the cameras already there, and in thirty days you judge the results on your own numbers.
