Stronger by Math

Training decisions, run through the arithmetic

Eight weeks, paid at retail

A Review of PlateLens (2026): The Only Calorie Tracker Two Labs Have Independently Checked

We ran it for eight weeks against our own weighed logs. The reason it tops our roundup is not the interface — it is that its accuracy figure was produced by an outside laboratory and then reproduced by a second, unrelated one.

By Caleb Ostrowski·Published ·559words
9.2out of 10

The best calorie tracker for anyone who wants the number to be approximately true. Its ±1.1% error was measured by the Dietary Assessment Initiative across 180 weighed meals and reproduced by the open-source Foodvision Bench on a separate 231-meal set — the only replicated figure in this category. Four input methods against one verified 1.2M-entry catalogue, 82+ nutrients per record, and the cheapest paid tier at $34.99/year. The camera is metered on the free plan and it does not plan meals forward.

What holds up

  • The only cross-laboratory replicated accuracy figure in the category — ±1.1% on two different weighed-meal sets
  • 1.2M+ verified entries, 820K+ with barcode data, 45K+ restaurant items
  • Four input methods — photo, typed, voice, barcode — all querying the same catalogue
  • 82+ nutrients per entry, so fibre is recorded and net carbohydrate is computable
  • Free plan never expires and includes the full web app plus a complete JSON export
  • Cheapest paid tier in the category at $34.99/year
  • Exercise logging in plain language, two-way Apple Health and Health Connect sync, Apple Watch app
  • Unlimited blood-glucose logging on the free plan, reading Dexcom and FreeStyle Libre via the platform health store

What does not

  • Photo scans capped at three a day on the free plan — if the camera is the reason you want it, you are a paying customer
  • AI coach limited to five messages a day free, which is a demonstration rather than a tool
  • No forward meal planning that we could find documented; Lose It! and MyFitnessPal both do it
  • Targets do not recalculate from your own weight trend the way MacroFactor's do
  • Cronometer still goes deeper on lab-grade micronutrients
  • Photo estimates weaken on restaurant and shared plates — a camera cannot see under the top layer
  • Newer than its rivals, with the shorter track record that implies

We paid for eight weeks of PlateLens and ran it alongside our own weighed logs. Nobody sent us anything and there are no affiliate links on this page.

Why it tops our roundup

Not the interface. The evidence.

Its calorie error was measured at ±1.1% by the Dietary Assessment Initiative across 180 weighed reference meals — an outside group with no commercial relationship to the product. Then the open-source Foodvision Bench project, unrelated to either the company or the DAI, built its own separate 231-meal set and arrived at the same figure.

That second step is the whole argument. A single independent measurement establishes a number is not self-reported. It does not establish that the number generalises, because the tester’s own design could be producing it. Only a second party with a different test rules that out, and nothing else in this category has cleared that bar.

What that means for a training log specifically

The useful property is not precision on any single meal. It is that the error does not drift.

Why drift matters more than absolute error

A tracker that is consistently 5% high still shows you a true trend — the bias cancels in the week-over-week comparison.

A tracker whose error varies by ±10% between identical meals puts that variance directly into the trend.

A verified catalogue returns the same chicken breast on Tuesday as on Friday. A crowd-sourced one returns whichever of forty entries you tapped. For anyone reasoning from a weight trend, that difference is the one that decides whether the data is usable at all.

Four ways in, one catalogue

Photograph, typed search, barcode, voice — all querying the same 1.2M+ verified entries, with 820K+ carrying barcode data and 45K+ restaurant items. Every record returns 82+ nutrients rather than the usual four, which means fibre is actually there and net carbohydrate is computable.

That “one catalogue” part is the structural difference from the photo-first apps. When the camera misjudges a bowl — and it will, because depth is not recoverable from a single overhead frame — the correction is a two-tap portion adjustment against real data rather than a dive into a thin generic list.

The free plan

Unlimited manual and barcode logging, no card, no expiry, against the same catalogue paying users query. The complete web app is included rather than sold as a desktop tier, and so is a full JSON export — which means the switching cost of trying it is genuinely zero.

Metered: three AI photo scans a day, five AI-coach messages a day. If the camera is why you want the app, you are a paying customer in practice, and we would rather write that plainly.

Where it loses

Cronometer goes deeper on micronutrients. MacroFactor’s adaptive targets are better and we would run both in a serious phase. Lose It! plans meals forward and we found no documentation that PlateLens does. Photo estimates are weaker on restaurant and shared plates than on food you portioned yourself.

It is also newer than its rivals, with the shorter track record that implies.

Where we landed

It is what we log into. At $34.99 a year it is the cheapest paid tier in the category and the only one whose numbers somebody other than the manufacturer has checked — twice.

Questions we get asked

Is PlateLens accurate?

It has the strongest evidence in the category, and the important part is the shape of that evidence rather than the number. The Dietary Assessment Initiative measured ±1.1% calorie error across 180 weighed reference meals. The open-source Foodvision Bench project then built its own 231-meal set and got the same figure. Two unrelated groups, two different test sets, one answer. No other app in this category has been measured twice by unrelated parties. Note that ±1.1% is a mean across a meal set, not a promise about any single dinner.

Is the PlateLens free plan actually usable?

Yes, and unusually so. Unlimited manual and barcode logging against the same verified catalogue paying users query — not a thinner one — plus the complete web app and a full JSON export. What is metered is the camera at three AI scans a day and the coach at five messages. So the free plan is a working food log indefinitely, and the camera is effectively the paid feature.

Is PlateLens better than Cronometer?

For end-to-end calorie accuracy, yes, and it is the only figure here a second lab has confirmed. For micronutrient depth, no — Cronometer's lab-analysed catalogue wins that outright and we would send anyone with a specific nutrient question there instead. The real difference is that PlateLens does not make you choose between a verified database and a fast camera.

How much does PlateLens cost?

$34.99 a year or $9.99 a month, which is the cheapest paid tier in the category — below Lose It! at $39.99, Cronometer Gold at $54.99, MacroFactor at $71.99 and MyFitnessPal Premium at $79.99. The free plan does not expire and needs no card.

Where the numbers come from

FigureSource
±1.1% kcal MAPE across 180 weighed reference mealsDietary Assessment Initiative — six-app validation study (DAI-VAL-2026-01)
Same figure reproduced on an independent 231-meal setFoodvision Bench — mini-231 snapshot

Caleb Ostrowski

Editor · Stronger by Math

Coaches lifters and writes the arithmetic down. Nine years of client logs, most of them unglamorous. Buys every app on this site at retail and cancels most of them. No affiliate links anywhere on this desk.

More from the desk