Not every measurement is exotic physics. Some of the most useful ones are pure clerical work: how many, how big, what does the dial say. These are jobs you already do — badly — with a squint and a mental note that evaporates by dinner. Three of Sentio's camera tools exist for exactly this: counting, measuring and reading, so the number lands in a catalog instead of a memory.
They're less glamorous than listening to a melon's resonance, and that's the point. A measurement platform that only handles the interesting cases isn't a platform; it's a party trick. The clerical tools are what make the catalog the place where all the numbers about your things live.
Count: how many are actually in there?
The Count tool counts distinct objects in view using on-device object detection. Spread the items so they're individually visible, capture, and get a number.
Humans are remarkable at estimating quantity and remarkably bad at exact counts past about a dozen — attention drifts, items get counted twice, someone talks to you at item thirty-one. The camera doesn't lose its place. The use cases are anywhere "roughly a lot" isn't good enough:
- Screws in a tray. The flat-pack manual says 48 of the small ones. Are there 48? Now you know before step 19, not during it.
- Bottles in a crate. Receiving a delivery, doing a stock check, splitting a bulk buy — count at a glance and the count is recorded, with a timestamp, under the object it belongs to.
- Plants in a bed. Seedlings germinated versus seeds sown is a germination rate — if anyone actually counts. Repeated counts of the same bed, filed together, turn into a record of what came up and what the slugs took.
Like every tool in the platform, detection runs on device. The frame is analyzed on your phone; what persists is the measurement.
Dimension: a payment card is a ruler you already carry
The Dimension tool measures an object's real size from a photo — with one elegant precondition. Photograph the object beside a reference of known size — a payment card or an A4 sheet — mark both on the photo, and read the object's real size off the ratio.
Why the reference? Because it makes calibration unnecessary. A photo alone can't tell you size: a matchbox close up and a shipping container far away can span the same pixels. Recovering real dimensions from raw pixels would mean knowing the camera's focal length, the sensor geometry and the exact distance to the subject — a per-device, per-shot calibration nightmare. But put a known object in the same photo, in the same plane, and every one of those unknowns cancels. Whatever the camera geometry did to the scene, it did equally to the card and to your object; the ratio between them survives untouched. A payment card is 85.6 mm wide everywhere on Earth — ISO standard, wallet-resident — and an A4 sheet is 210 × 297 mm. If your object spans 3.2 card-widths, it's about 274 mm, on any phone, with any lens, at any distance. This is the platform's relative-features principle in its most literal form: never trust an absolute pixel, always trust a ratio.
The limitation is the flip side of the same geometry, and it's worth stating plainly: the method works when both the object and the reference lie flat at the same distance from the camera. The cancellation only holds if perspective treats both identically. A card on the table next to a standing vase measures the vase's footprint honestly and its height not at all; a reference propped at an angle skews the ratio. Lay both flat, shoot from above, and the numbers are trustworthy — measuring furniture for a doorway, sizing a part you need to replace, recording how big that mystery flange actually is before you drive to the store.
Meter reading: the dial becomes a trend
The Meter reading tool reads a utility meter's register with the camera. Digits are recognized on device, you confirm the value and unit, and the reading files under that meter's object in your catalog.
That confirmation step is deliberate, so let's dwell on it. This is assistance, not blind OCR trust. Meter registers are hostile territory for character recognition — dirty glass, half-rolled digits mid-transition, reflections, that one wheel that's been misaligned since 1998. A tool that silently commits whatever it thought it saw would poison your records with the occasional confident misread, and a consumption record with silent errors is worse than no record. So the recognizer does the tedious part — finding and reading the register — and shows you what it saw; you glance, correct if needed, confirm. Ten seconds instead of transcribing digits onto a scrap of paper, with your eyes as the final authority.
The design rule: when a reading feeds a permanent record, the human confirms it. On-device recognition does the labor; the user does the vouching. It's the same honesty that runs through the platform — a tool should make errors visible and correctable at capture time, not launder them into the catalog as facts.
And here's where the catalog stops being filing and starts being measurement: repeated readings of the same meter object become a consumption trend. One reading is a number; a series is a slope. Kilowatt-hours per day, cubic meters per week — the differences between confirmed readings, over time, are the measurement you actually wanted. Did the new heat-pump settings help? Does consumption spike when the in-laws visit? Is the water usage creeping in a way that smells like a slow leak? A meter is read the same way whether you rent or own, whatever the utility's app does or doesn't show you, and the trend lives on your device.
Small tools, one catalog
None of these three will headline a keynote. But notice what they share: each replaces a moment of squinting with a record — a count with a timestamp, a size with a stated method, a meter value you vouched for. Each files under an object, so the numbers accumulate where they belong: this parts tray, this doorway project, this gas meter. And each measurement can take a ground-truth label after the fact — the count verified during assembly, the dimension checked with a tape measure once you found one — so the tools' accuracy is itself measured, on real scenes, over time.
The catalog, and everything in it, works fully offline — the how and why of that is its own story. The camera's more physical modes — color, brightness, flicker, rotation — are covered in Color and light and Blinking, flicker and rotation. For the whole map of what a phone can measure, start with the tools overview.