Repolode
Submit
rdumasia303

rdumasia303/deepseek_ocr_app

Web app for running DeepSeek-OCR on images and PDFs, with a React frontend, FastAPI backend and Docker setup.

Open-source alternative to:Adobe Acrobat

View on GitHub
Stars
1,892
Forks
311
Open issues
12
Last commit
7 months ago

What is it?

DeepSeek OCR App wraps the DeepSeek-OCR model in a web interface. A React frontend lets you upload images or PDFs, pick an OCR mode such as plain text extraction, description, term finding or custom prompts, and view the results. A FastAPI backend runs the model and exposes endpoints for single images and multi-page PDFs.

For PDFs, the app processes each page with progress tracking and can export the result as Markdown, HTML, DOCX or JSON. Embedded images can be extracted and included in the output, and LaTeX formulas and tables are preserved where the model produces them. For image mode, grounding bounding boxes are drawn over the picture.

The whole stack runs through Docker Compose and needs an NVIDIA GPU with CUDA support, roughly 8-12GB of VRAM, and about 5-10GB of model download on first start.

Key features

  • 01Single image OCR with plain text, description, term finding and freeform prompt modes.
  • 02Multi-page PDF processing with per-page progress tracking.
  • 03Export to Markdown, HTML, DOCX or JSON.
  • 04Extracts embedded images from PDFs and includes them in exported documents.
  • 05Preserves LaTeX formulas and tables produced by the model.
  • 06Runs via Docker Compose with a configurable .env file.

Use cases

  • Converting scanned PDFs into editable Word or Markdown files.
  • Pulling tables and text out of invoices, forms or reports.
  • Extracting text and formulas from academic papers.
  • Testing DeepSeek-OCR locally through a web UI and REST API.

Quick start

  1. 1

    Clone the repository and copy the environment file

    git clone <repository-url> cd deepseek_ocr_app cp .env.example .env
  2. 2

    Start the application with Docker Compose

    docker compose up --build
  3. 3

    Open the frontend in your browser at localhost:3000

Things to know

  • Requires an NVIDIA GPU with CUDA support and the NVIDIA Container Toolkit.
  • The first run downloads a model of roughly 5-10GB and needs about 20GB free disk space.
  • Advanced OCR modes such as tables, layout and multilingual are disabled pending further testing.

Topics

#ocr#deepseek#pdf#document-conversion#fastapi#react#docker#computer-vision

FAQ

What is deepseek_ocr_app?
Web app for running DeepSeek-OCR on images and PDFs, with a React frontend, FastAPI backend and Docker setup. DeepSeek OCR App wraps the DeepSeek-OCR model in a web interface. A React frontend lets you upload images or PDFs, pick an OCR mode such as plain text extraction, description, term finding or custom prompts, and view the results. A FastAPI backend runs the model and exposes endpoints for single images and multi-page PDFs.
Is deepseek_ocr_app free to use? What license does it have?
deepseek_ocr_app is open source under the MIT license. Check the license terms before using it in commercial products.
Is deepseek_ocr_app still maintained?
No commits for 6+ months. The last commit to deepseek_ocr_app was on Mar 31, 2026.
What can I use deepseek_ocr_app for?
Converting scanned PDFs into editable Word or Markdown files. Pulling tables and text out of invoices, forms or reports. Extracting text and formulas from academic papers. Testing DeepSeek-OCR locally through a web UI and REST API.
How do I install deepseek_ocr_app?
Clone the repository and copy the environment file: git clone <repository-url> cd deepseek_ocr_app cp .env.example .env Start the application with Docker Compose: docker compose up --build Open the frontend in your browser at localhost:3000
Where can I get deepseek_ocr_app?
The source code and installation instructions are on GitHub: https://github.com/rdumasia303/deepseek_ocr_app

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects