README.md

metadata

language:
  - tr
  - en
license: mit
task_categories:
  - image-classification
  - feature-extraction
  - zero-shot-image-classification
tags:
  - food
  - nutrition
  - vision
  - calorie-tracking
  - turkey
size_categories:
  - 10K<n<100K

Turkish Food & Nutrition Vision Dataset (Food v2)

High-resolution, human-POV Turkish food and packaged snack dataset with canonical nutrition macros and portion sizes.

📊 Dataset Summary

• Repository: ozertuu/foodv2
• Canonical Foods: 29,840
• Images Total: 89,520
• Average Images / Food: ~3.0
• Image Sources: Yemeksepeti / DeliveryHero, GetirYemek, Migros Sanalmarket, Getir, Nefis Yemek Tarifleri, OpenFoodFacts.

🍽️ Features & Schema

• image: PIL Image / Raw JPEG bytes
• food_id: Unique canonical food UUID
• name_tr: Turkish food name
• name_en: English translation
• category: Macro food category (proteins, sweets, vegetables, fruits, grains, etc.)
• calories_100g, protein_100g, carbs_100g, fat_100g: Canonical macro nutrition per 100g
• default_serving, default_serving_grams: Standard reference portion
• source_url, source_domain: Source origin
• sha256, phash: Deduplication hashes

💻 Quickstart in Kaggle / Python

from datasets import load_dataset

# Load entire dataset or stream
dataset = load_dataset("ozertuu/foodv2", split="train")

# Inspect first item
item = dataset[0]
print(item["name_tr"], item["calories_100g"], "kcal")
item["image"].show()