Gramwise — Credits & licenses

Last updated: August 21, 2026

Everything Gramwise knows comes from open data, open models, and open source — listed here with their licenses. Questions: hollownett@gmail.com.

Nutrition data

• USDA FoodData Central — The core of the bundled food database (US) — CC0 1.0 (public domain)
• CIQUAL (ANSES) — French food composition table — Licence Ouverte 2.0 (Etalab)
• Fineli (THL) — Finnish food composition data — CC BY 4.0
• CoFID (UK OHID) — UK food composition data — Open Government Licence v3.0
• Frida (DTU) — Danish food composition data — Free with citation
• Open Food Facts — Barcode & product lookups, only when you tap "Search online" — ODbL 1.0

Recognition models & training data

• Open Images V7 — Food photos the built-in recognizer trained on — Annotations CC BY 4.0; images CC BY 2.0
• Grocery Store Dataset — Grocery item photos for training — MIT
• Vegetable Image Dataset — Vegetable photos for training — CC BY-SA 4.0
• Fruit and Vegetable Image Recognition — Produce photos for training — CC0 1.0 (public domain)
• Fruits-360 — Augmentation-only produce photos — CC BY-SA 4.0
• MyFoodRepo (AIcrowd) — Real-meal photos + the food-outline training data — CC BY 4.0 (AIcrowd Food Recognition)
• Nutrition5k (Google Research) — Dish photos with weighed nutrition, the portion estimator's training data — CC BY 4.0

Fonts & open source

• Instrument Sans — The app's typeface — SIL Open Font License 1.1
• JetBrains Mono — The numeric typeface — SIL Open Font License 1.1
• Ionicons — Every icon in the app — MIT
• llama.cpp / llama.rn — Runs the optional downloadable AI Pack models — MIT
• react-native-fast-tflite + TensorFlow Lite — Runs the built-in recognition model on your phone — MIT / Apache 2.0

Open-source license texts

The verbatim MIT, Apache-2.0, and SIL Open Font License texts for the components above travel with the app and are readable offline in Gramwise under Settings → Legal & credits → Credits & licenses.