Image to Text OCR API

A FastAPI-based OCR service that extracts text and structure from images, supporting advanced features like preprocessing, language selection, confidence scores, and more.

Features

• Extracts text from uploaded images or image URLs
• Returns detected languages with confidence
• Returns bounding boxes for blocks, paragraphs, lines, and words
• Includes confidence scores for each word
• Supports image preprocessing: grayscale, thresholding, denoising
• Allows language selection for OCR
• Can automatically deskew (rotate) images

Installation

1. Clone the repository and navigate to the project directory.
2. Install dependencies:

   pip install -r requirements.txt

3. Make sure Tesseract OCR is installed and in your system PATH.

Running the API

uvicorn main:app --reload

Or, if uvicorn is not in your PATH:

python -m uvicorn main:app --reload

API Usage

Visit http://localhost:8000/docs for interactive documentation.

Endpoint: /extract-text/ (POST)

Accepts either an image file upload or an image URL, with optional processing options.

Form Fields

• file: (optional) Image file to upload
• image_url: (optional) URL to an image
• preprocess: (optional) Comma-separated preprocessing steps: grayscale, threshold, denoise
• ocr_lang: (optional) Tesseract language(s), e.g. eng, fra, eng+fra (default: eng)
• deskew: (optional) true or false (default: false), deskew image before OCR

Example Request (using curl)

curl -X POST "http://localhost:8000/extract-text/" \
  -F "file=@your_image.png" \
  -F "preprocess=grayscale,threshold" \
  -F "ocr_lang=eng+fra" \
  -F "deskew=true"

Example Response

{
  "status": true,
  "text": "...extracted text...",
  "boxCoordinates": [0.1, 0.2, 0.3, 0.4],
  "blocks": [
    {
      "boxCoordinates": [ ... ],
      "paragraphs": [
        {
          "boxCoordinates": [ ... ],
          "lines": [
            {
              "boxCoordinates": [ ... ],
              "text": "...",
              "words": [
                { "text": "...", "boxCoordinates": [ ... ], "confidence": 95.0 }
              ]
            }
          ]
        }
      ]
    }
  ],
  "detectedLanguages": [
    { "languageCode": "en", "confidence": 0.98 }
  ],
  "executionTimeMS": 1234
}

Docker

A Dockerfile is provided for containerized deployment. Build and run with:

docker build -t image-to-text-api .
docker run -p 8000:8000 image-to-text-api

Notes

• For best results, ensure Tesseract is installed and available in your system PATH.
• You can combine preprocessing options for improved OCR accuracy.
• The API supports both file uploads and image URLs.

License

MIT