Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
veganism.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Veganism Social is a welcoming space on the internet for vegans to connect and engage with the broader decentralized social media community.

Administered by:

Server stats:

302
active users

veganism.social: AboutStatusPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.0-alpha.5+vegan

#ocr

9 posts8 participants1 post today
Public *
Nielso

Früher dachte ich, man bräuchte #OCR, um ältere Bücher in in die so leicht durchstöberbare digitale Welt rüberzuretten.

Was ich heute mit OCR mache: Ich drucke E-Mails in PDF-Dateien aus, in denen die benötigten Infos als etliche Screenshots/Pixelgrafiken eingebettet sind, und lasse dann #OCRmyPDF drüber laufen, um darin suchen zu können.

#Digitalisierung, so ein großes, mitunter wahlkämpferisches Wort, um damit Leute zu überzeugen, die nicht mal E-Mail sachdienlich benutzen können.

Public
GripNews

🌘 GitHub - ses4255/Versatile-OCR-Program：優化用於機器學習訓練的多模式 OCR 管道（文字、圖形、數學、表格、圖表）
➤ 專為機器學習訓練而生的多功能 OCR 管道
✤ github.com/ses4255/Versatile-O
這個GitHub專案「Versatile-OCR-Program」提供一個針對機器學習訓練優化的多模式 OCR 管道。它能夠處理包含文字、圖形、數學公式、表格和圖表的複雜教育材料，並提供結構化輸出，例如JSON或Markdown格式，方便模型訓練。系統支援多種語言（日文、韓文、英文）並具有高準確率，尤其適用於學術資料集。
+ 這個專案對於需要處理大量學術論文或考試題目的研究人員來說，簡直是救星！它能自動提取並結構化資料，省去了大量人工處理的時間。
+ 這套系統不僅支援多種語言，而且對於數學公式和圖表的處理能力也相當出色，讓我對它在教育領域的應用充滿期待。
#機器學習 #OCR #開源專案

Public
GripNews

🌘 Omni OCR 基準測試
➤ 評估大型語言模型 OCR 能力的開源工具
✤ github.com/getomni-ai/benchmar
本文件描述了 Omni OCR 基準測試工具，用於評估不同大型多模態模型（如 gpt-4o）的 OCR 和資料提取能力。該基準測試比較了傳統 OCR 供應商和語言模型的 OCR 準確性，並提供開源的評估資料集和方法。主要評估指標為 JSON 準確性和文字相似度，並提供運行基準測試的詳細步驟和支援的模型清單，包含閉源和開源 LLM 以及雲端 OCR 供應商。使用者可以透過設定模型參數和 API 金鑰來運行測試，並查看結果。
+ 這個基準測試對研究 OCR 和 LLM 在資料提取方面的應用非常有幫助，能更客觀地比較不同模型的優劣。
+ 開源的評估方式讓人很放心，可以根據自己的需求擴展測試範圍，參考價值很高。
#人工智慧 #OCR #基準測試 #大型語言模型

Public *
Aral Balkan

So if you’re using Mastodon on the web, you can press the ⚠️ALT button and then follow the “Detect text from picture” link.

On Mac/iOS, you can select text on images as if they were text by clicking/tapping and dragging and paste that in (might be more accurate; that’s what I did).

PS. This was meant to be a reply to mastodon.social/@fatbrit/11421 but somehow didn’t get threaded correctly (was using the web client instead of Mona. I somehow manage to do that there sometimes. Has happened before.) :)

Screenshot of Mastodon posting interface showing a picture being added with the ALT button and Detect text from picture link highlighted.
ALT
#altText#fediTips#detectFromPicture
Public
Richard Audette

Trying some modern OCR tools recently: marker - github.com/VikParuchuri/marker and Mistral OCR - mistral.ai/news/mistral-ocr

I last looked at this in 2015, when James Bond entered the public domain: hotelexistence.ca/james-bond-e . With 2015 tools, the OCRed output of Bond books was poor.

The new generation is better, but still requires human review.

I admire the quality of work done by Project Gutenberg in their creation of digital editions of books in the public domain.

Output of Mistral AI's 2025 OCR vs a package bundled with a scanner in 2015
ALT
#projectgutenberg#ocr
Public
Jesus Castagnetto 🇵🇪

And ultra compact #AI model for document #OCR by #IBM and #HuggingFace

marktechpost.com/2025/03/18/ib

MarkTechPost · IBM and Hugging Face Researchers Release SmolDocling: A 256M Open-Source Vision Language Model for Complete Document OCRConverting complex documents into structured data has long posed significant challenges in the field of computer science. Traditional approaches, involving ensemble systems or very large foundational models, often encounter substantial hurdles such as difficulty in fine-tuning, generalization issues, hallucinations, and high computational costs. Ensemble systems, though efficient for specific tasks, frequently fail to generalize due to their dependency on handcrafted pipelines for each sub-task. On the other hand, multimodal foundational models, although powerful, often suffer from high computational costs and reliability issues like hallucinations. Researchers from IBM and Hugging Face have recently addressed these challenges by releasing SmolDocling, a 256M
VegansExploreLive feeds
About