Sam Wilson on Nostr: There's a new version of Wikimedia OCR out now (1.8.0), with some new localisation ...
There's a new version of Wikimedia OCR out now (1.8.0), with some new localisation strings and the addition of Northern Min script, and interestingly we've switched to using the document-text-detection method of Google's API, rather than just text-detection. But we already did that change a few years ago! Oops. Somehow it was changed back. The non-document variant is for when you also want to support things like finding signs in photos.
https://ocr.wmcloud.org/Published at
2026-03-02 13:39:00 UTCEvent JSON
{
"id": "f01bb6b25fbc6b5c5b269120b3f1d146521c61e5e2aa779474e3053cd4491d41",
"pubkey": "8ad46dd46db2fec4b1f0344b6daaeabb618e3e6d979c30b278cc04dd3f3078f4",
"created_at": 1772458740,
"kind": 1,
"tags": [
[
"proxy",
"https://wikis.world/users/samwilson/statuses/116159855993318784",
"activitypub"
],
[
"client",
"Mostr",
"31990:6be38f8c63df7dbf84db7ec4a6e6fbbd8d19dca3b980efad18585c46f04b26f9:mostr",
"wss://relay.ditto.pub"
]
],
"content": "There's a new version of Wikimedia OCR out now (1.8.0), with some new localisation strings and the addition of Northern Min script, and interestingly we've switched to using the document-text-detection method of Google's API, rather than just text-detection. But we already did that change a few years ago! Oops. Somehow it was changed back. The non-document variant is for when you also want to support things like finding signs in photos.\n\nhttps://ocr.wmcloud.org/",
"sig": "0c48167866e22a65bafdac040be6187c0e76d1e3ec0478a3e65473a72443c2f8ecd5ce4e056bc661e8e59a0aba367b78decf0c4b397c8d6285a7435139c8b404"
}