Skip to content
LIVENew: Complete workflows from web or mobile.Try it free →

Tiliter Vision Agent

Object Counting Agent

Count specific objects from images automatically using Tiliter Vision Agents. Ideal for inventory, proof of delivery, traffic analysis, tools, assets and operational workflows.

Object Counter

Stop counting stock by hand.

Tiliter counts items in a photo the moment it's submitted – accurately, instantly, and without the recounts. From tyres and pallets to parts and packages.

Submitted photoAnalysing…

Tyre yard · Inventory check

Yard 03 · Bay A · 09:18 · Daniel R.

Tyres stacked in rows inside a delivery truck, ready to be counted
Result
0
tyres counted
AI confidence0%

Detection summary

  • Items detected67
  • Confidence99.4%
  • Logged to inventoryYes
Counted in10s
Manual count + recount~3 min

What this means for your operation

Faster counts. Cleaner inventory data.

Illustrative impact for a team running ~400 stock counts per day across sites

10sper countvs. ~3 minutes manually
99.4%count accuracyno missed items, no recounts
95hours saved / weekacross the operation
$142ksaved per yearin stocktake labour alone

Real-World Counting Use Cases

Use the Object Counting Agent to automate counting tasks, reduce manual effort and improve operational visibility across industries.

Products & Inventory

Count products on shelves, in packaging, displays or storage locations.

Tools, Parts & Equipment

Verify required tools, components or safety gear before and after tasks.

Deliveries & Materials

Count pipes, boxes, pallets, parts or delivered goods quickly and accurately.

People & Vehicles

Track traffic flow, occupancy, queues or vehicle movement across sites.

Built for real counting workflows

The Object Counting Agent fits into existing operations by automatically counting visible items in submitted images and returning structured results for review, reporting or workflow automation.

Capture item evidence

Teams submit images from mobile devices, inspection workflows, warehouse processes, cameras or existing operational systems.

AI counts the objects

The Object Counting Agent identifies and counts visible items such as products, parts, packages, assets, tools or other defined objects.

Results move into workflows

Counts can be reviewed, compared against expected quantities, sent into reports or used to trigger follow-up actions.

Object Counter questions, answered

What can the Object Counting Agent count?

Anything visually distinct enough to tell apart in a photo: tyres, pallets, cartons, packages, tools, parts, bottles, crates, plants. It counts a defined object type rather than everything in frame, so you get “34 cartons” rather than a total of every shape the camera saw.

How fast is a count?

It depends on how much is in the frame. A single item comes back in a few hundred milliseconds. A full rack or pallet with many items to separate takes up to about ten seconds, against roughly three minutes to count the same rack by hand and longer again if it needs a recount. The consistency matters more than the speed: the same photo counted twice returns the same number.

How accurate is AI object counting?

Every count returns a confidence score alongside the number, so low-confidence results can be routed for a human check instead of being trusted silently. Accuracy depends on your images, since lighting, angle and occlusion all matter, which is why the confidence score is part of the output rather than hidden behind it.

Can we count from a phone photo, or do we need fixed cameras?

Either. Teams capture on web or mobile as part of a workflow, and counts can also run against images from fixed cameras. The CCTV and IP camera integration guide covers feeding existing camera infrastructure through the API.

Do we have to train it on our own products first?

No. Counting runs on what you submit, with no training project first. If you need specific items identified rather than counted, that is product recognition, and much of that works as it is too, with a catalogue needed only where the answer has to be your exact SKU.

What is the difference between object counting and product recognition?

Counting answers “how many of this thing are here”. Product recognition answers “which exact item is this”, identifying a specific SKU or variety. Stocktaking usually wants counting; checkout and inventory accuracy usually want recognition. Many operations use both.

Can counts feed into the systems we already use?

Yes, and normally by configuring what you already run rather than building against anything. A count can land on the record it belongs to – in Salesforce, for instance – or be pushed to whoever needs to act on it through Slack, Zapier or Power Automate. Nobody has to remember to go and look somewhere.

What are people using it for?

Stocktakes and cycle counts set up as recurring workflows, proof of delivery verification, yard and asset counts, and counting parts or tools before and after a job. The pattern is the same each time: the count is assigned to a team, captured in the app where the work happens, and kept as a record instead of a number written on a clipboard and typed in later.

Ready to see the Object Counter in action?

Build visual verification workflows and deploy them through mobile, web, cameras or your existing systems. Deployed in thousands of locations worldwide and ready to scale with your operations.