今回はOCR（PDFや画像データの文字認識）用ライブラリを紹介します。OCR用のサンプルデータは下記の通りです。 シンプルな読み込みはtabula.read_pdf(filepath, pages='all')とします。またfilepathにurlを指定すればweb経由で取得も可能です。 下記の通り戻り値はリスト ...

In daily office work and development, we often need to extract text from specific regions of a large number of PDF files (e.g., dates/amounts on invoices, key indicators on reports) or capture ...

Standalone tool that adds native text to scanned/image-only PDFs using DocTR word detection + Qwen VLM text recognition. pdf-native-ocr/ ├── run.py # CLI entry point (config at top) ├── api.py # ...

I want to share the approach, the trade‑offs, and the architectural decisions behind it — in case you’re building something similar or exploring how lightweight OCR can fit into a modern distributed ...