About MyFoodMate AI

How we calculate macros

The model recognizes the dish and its components from a photo and estimates the weight of each from visible area and typical serving sizes. Calories and macros then come not “from the model” but from reference food composition tables - so the result is reproducible and you can check it.

The key difference from Western trackers: we count post-Soviet dishes by component, not as one averaged entry. Borscht is broth, vegetables, meat and sour cream, and you can correct each component separately.

Data sources

Straight talk about recognition accuracy

We don't claim “99% accuracy” - no photo-based count can deliver that. The model sees a snapshot, not a scale: some weights are estimates, and the margin depends on how clearly the portion is visible.

That's why the result always comes with a range: ±10 kcal for a packaged product, around ±45 for a simple home-cooked dish, up to ±175 when the portion is hard to see. Confirm the weight and the range roughly halves. That's accurate enough for tracking weekly averages, but not for medical record-keeping - and we say so plainly.

Where we stand on nutrition

We don't promote crash diets, detoxes or “lose 10 kg in a month” schemes, and we never will. The service is built around balance and weekly averages: one day of overeating changes nothing, while the habit of logging your food changes almost everything.

• We don't split food into “good” and “bad”FoodMate Score shows what's missing instead of banning foods.
• We never suggest a deficit above 20%A safe rate of weight loss is 0.3–0.7 kg per week.
• We don't use stigmatizing languageNo “you broke your limit” and no judgements about bodies.
• We don't replace your doctorWith a medical condition or a prescribed diet, follow your specialist's advice.

Company details and contacts