Why this is different

Most calorie apps assume a Western plate.
Calorite doesn't.

Ask a typical AI photo-scan model to estimate a bowl of dal, and it will often strip out the ghee, guess a Western-sized portion, and price the dish in USDA terms alone — because that's the plate its training data was built around. Calorite classifies the cuisine first, then applies that cuisine's own conventions.

Every photo scan runs through a cuisine classification step before estimation begins. The model looks at the tableware, plating, and ingredients and assigns one of four buckets — then applies rules specific to that bucket, instead of one universal assumption.

南

South Asian

Dal, thali, biryani, dosa, idli, curry
The typical cooking oil or ghee of the dish, as normally prepared, is assumed by default — not a "fat-free curry" guess. Nutrition is cross-checked against the Indian Food Composition Tables (IFCT), not USDA alone.
東

East Asian

Ramen, stir-fry, sushi, fried rice, noodle soups
Cooking oil and sauce levels are estimated to match how the dish is actually prepared, rather than zeroed out by a generic vision model.
W

Western

Roast dinners, burgers, pasta, salads, sandwiches
Standard USDA FoodData Central portion and macro baselines — the case most calorie apps already handle reasonably well.
+

Everything else

Tacos, mezze, packaged and branded products, and more
Backed by a growing, moderated Global Food Database of community-submitted dishes for foods neither reference database covers well.
Two reference databases, not one

USDA alone isn't built for every kitchen.

USDA FoodData Central is the default reference most US-built calorie apps rely on — it's comprehensive for American and Western food, but thin on South Asian home cooking. Calorite cross-checks against the Indian Food Composition Tables (IFCT) as well, so a dal or a dosa isn't approximated from the nearest Western analog.

Hidden-fat detection. Cooking oil and ghee are easy for a vision model to miss entirely, especially in a photo where the oil has mostly been absorbed into the dish. When the model sees a real risk of extra fat beyond what's typical — pooling oil, a heavy shine, visible ghee, a deep-fried surface — it flags it, and the app offers one-tap chips to add the specific oil or sauce yourself, rather than silently under-counting.

Community-extended

The Global Food Database

No fixed reference database covers every regional dish, home recipe, or regional packaged product. Calorite's Global Food Database lets users submit real dishes — moderated before they're searchable — so coverage keeps growing in the direction actual users are logging, not just where a corpus happened to be built.

See it on your own plate

Scan a meal that isn't a burger.

Start a free trial and try the photo scan on whatever you're actually eating.