| --- |
| title: EAN Data |
| --- |
| |
| # EAN Data |
|
|
| Open barcode data for Denmark. |
|
|
| A barcode is the one thing a product carries that a machine can read with |
| certainty. If you know what a number means, software can help someone decide |
| what to do with the item in front of them. That lookup does not exist as open |
| data for the Danish market, so we are building it. |
|
|
| ## What we publish |
|
|
| **[ean-product-dataset](https://huggingface.co/datasets/EAN-data/ean-product-dataset)** |
| — barcode to product name, description, product image, retailer and safety |
| data sheet link, harvested from public product pages at Danish retailers. |
|
|
| | | | |
| |---|---| |
| | Products | 1,132,080 | |
| | Retailers | 153 | |
| | With a fetched image | 909,774 | |
| | With a safety data sheet link | 31,773 | |
| | Assessed as chemical products | 56,461 | |
|
|
| ## How we work |
|
|
| **Images are stored as original bytes, not as links.** A URL is a promise: a |
| retailer can change CDN tomorrow, and the dataset would be empty without |
| looking empty. Every image is kept byte for byte as the retailer sent it — |
| same format, same resolution. Re-encoding and downscaling are both |
| irreversible, and a downscaled JPEG is too little if a model is later to read |
| the label text. |
|
|
| **Every barcode is verified with its check digit,** and only counted when the |
| page itself labels it as an EAN, GTIN or barcode. A loose 13-digit number is |
| just as often a phone number or a form placeholder. |
|
|
| **"Not assessed" is not "not hazardous".** `is_chemical` is `true`, `false` |
| or `null`, and `null` means nobody has looked yet. Filtering that distinction |
| away would be a claim we cannot support. |
|
|
| **Stock photos are labelled, not removed.** When the same image file appears |
| on 25 or more different items, it is the shop's "image to follow" icon rather |
| than a product photo — we mark it and leave it in. A silent filter is a claim |
| nobody can check. |
|
|
| ## Why |
|
|
| The dataset grew out of work on chemical waste sorting: staff at Danish |
| recycling centres photograph a product and are told which container it |
| belongs in. Getting that right needs to know what the barcode is, and the |
| lookup turned out to be the hard part. It seemed more useful shared than kept. |
|
|
| ## Provenance and rights |
|
|
| Content was fetched from retailers' public product pages; every row records |
| its `source`, `product_url` and `harvested_at`. |
|
|
| Product names, descriptions and photographs remain the property of the |
| respective retailers and manufacturers. We claim no rights over them and |
| publish the data as-is, for research and for building tools that help people |
| sort chemical waste correctly. If you are a rights holder and want a row or |
| an image removed, open a discussion on the dataset and it will be removed. |
|
|
