Ingredient Atlas

Open ingredient and household catalog images for recipe, grocery, pantry, and meal-planning apps.

I made this because recipe apps usually need a simple thing that is weirdly annoying to find: a clean image for garlic, lemon juice, cassava flour, frozen broccoli, and hundreds of other ingredients, with stable slugs and metadata attached.

Ingredient Atlas is incubated by Beets, but it is app-agnostic. You can use it without Beets, private services, or an API key.

Links:

• Dataset: https://huggingface.co/datasets/ionicam/ingredient-atlas
• npm: https://www.npmjs.com/package/ingredient-atlas
• Issues: https://github.com/ionmesca/ingredient-atlas/issues

What You Get

• 1,830 public catalog records
• 5,490 public image files
• food, household, personal-care, and pet shopping item coverage
• WebP thumbnails and PNG fallbacks
• stable slugs, aliases, categories, checksums, and review status
• JSONL, Parquet, full manifest, and compact manifest
• public-safe metadata with internal IDs and prompts redacted

Quick Use

Install the tiny resolver package:

npm install ingredient-atlas

Then resolve an ingredient to the public Hugging Face image files:

import { getIngredientImage } from "ingredient-atlas"

const garlic = getIngredientImage("garlic", {
  baseUrl: "https://huggingface.co/datasets/ionicam/ingredient-atlas/resolve/main",
})

console.log(garlic.url)

Or use the dataset files directly from Hugging Face:

https://huggingface.co/datasets/ionicam/ingredient-atlas

Catalog Helpers

The resolver has catalog-named helpers for food, household, personal-care, and pet shopping items:

import { getCatalogItemImage } from "ingredient-atlas"

const garlic = getCatalogItemImage("garlic", { kind: "food" })

Why This Is Different

Most food image datasets are for model training, dish classification, product labels, or nutrition research. Ingredient Atlas is for app builders. The images are isolated catalog assets with metadata that is useful in UI.

License

• Code: MIT
• Metadata: CC0-1.0
• AI-generated images: CC0-1.0

Release approvals are tracked in docs/PUBLISHING.md.

AI And Nutrition Notes

Images are AI-generated and reviewed on a best-effort basis. They are useful, not perfect.

Nutrition metadata is best-effort ingredient metadata. Some values are USDA-backed, while others are approximate or missing. Do not use it as medical, allergy, or dietary advice.

Corrections

Found a wrong image, wrong form, bad alias, or missing ingredient? Open a GitHub issue or email [email protected].

Relationship To Beets

Ingredient Atlas was incubated by Beets. Beets supplied the starting taxonomy, image generation workflow, and review pipeline. Ingredient Atlas is meant to be useful to anyone building food, grocery, pantry, and household planning software.