Cuisine Classification
Tag every menu item with its cuisine across 19 classes and 100+ languages. The model says so plainly when it is uncertain.
Paste your own menu, one item per line. The classifier tags each item with its cuisine.
The playground classifies up to 40 items per run. The API itself handles 512 per call.
Verification is not set up here, so live runs are paused. You can still browse the examples.
The classifier covers 19 cuisine classes and is strong on most of them. Korean and Southeast Asian are its weaker classes on average, where harder or less common dishes are mislabeled more often or come back uncertain. Macro F1 is 0.77 across all classes.
Source: classifier v5 evaluation
- Margherita PizzaItalianconfidence97%0.968
- Tonkotsu RamenJapaneseconfidence61%0.607
- Tacos al PastorMexicanconfidence97%0.972
- Tom Yum SoupThaiconfidence98%0.981
- Mapo TofuChineseconfidence77%0.774
- Butter ChickenNorth Indianconfidence86%0.864
- Masala DosaSouth Indianconfidence100%1.000
- BibimbapKoreanconfidence94%0.938
- Pho BoSE Asianconfidence87%0.874
- Chicken ShawarmaMiddle Easternconfidence87%0.869
- Coq au VinEuropeanconfidence95%0.950
- Mac and CheeseAmericanconfidence67%0.674
The bar is the classifier confidence. An uncertain item gets a dashed-border pill, the model declining to commit rather than guessing.
Use this in your app
Pre-filled with the items you just classified. Drop in a key and run it.
curl -X POST https://dish-embed.latimal.com/classify \
-H "X-API-Key: $LATIMAL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"items": [
"Margherita Pizza",
"Tonkotsu Ramen",
"Tacos al Pastor",
"Tom Yum Soup",
"Mapo Tofu",
"Butter Chicken",
"Masala Dosa",
"Bibimbap",
"Pho Bo",
"Chicken Shawarma",
"Coq au Vin",
"Mac and Cheese"
]
}'Classify your real catalog
This menu is curated. The honest test is your own catalog. Get a free key with 5,000 credits, or send your menu to the founder.