No black box. Here’s exactly how PlateLens calculates every calorie, macro, and micronutrient.
Photograph the plate and an AI vision model — trained on 4.2 million labeled food images across 12,000+ food categories from over 100 countries — identifies each item. Type or dictate a meal and the same identification runs from your words. Barcodes skip straight to the product database.
From a photo, a proprietary depth-estimation algorithm calculates volume and weight using plate geometry, food density profiles, and reference object detection — no props or scale needed. From text or voice, portions come from the amounts you say — “150 g”, “a cup” — and every portion stays editable.
Identified items are matched against our multi-source nutrition database, verified for quality before inclusion:
Confidence scores are assigned at every step. When data sources disagree or coverage is limited, PlateLens shows you ranges instead of false precision — because honest estimates beat confident guesses.
Barcodes are looked up against a mirror of Open Food Facts, the open product database, with Nutritionix extending branded coverage.
The database is updated weekly; new products are indexed within 48 hours of appearing in partner databases. You never search a database by hand: describe the meal — or photograph it, or scan it — and PlateLens does the lookup, and every ingredient and portion stays editable after logging.
The same analysis records each drink’s alcohol strength (ABV) and the liquid you drank. For a single drink, pure alcohol is milliliters × ABV ÷ 100 × 0.789 g/ml, so a 500 ml beer at 5% has 19.7 g. For a mixed drink, it is the energy that protein, carbohydrates and fat do not explain, divided by 7 kcal per gram, once for the whole drink. A barcode uses the product’s declared % vol. When a drink’s numbers do not agree, it reads No estimate instead of a guessed amount. Alcohol energy is already part of the calories. See what counts as alcohol.
See also: portions and corrections.