WriteVision: Handwriting Recognition using TrOCR

WriteVision is a project that leverages the TrOCR (Transformer-based Optical Character Recognition) model from Microsoft to recognize handwritten text from images. Built on the powerful Transformer architecture, TrOCR excels at complex image-to-text tasks, making it ideal for handwriting recognition.

Features Handwritten Text Recognition: Accurately extracts text from images (supports URLs, local files, and Base64 strings).

Pre-trained Model: Utilizes the microsoft/trocr-base-handwritten model from Hugging Face, trained on the IAM Handwriting dataset.

Extensible: Easily adaptable for different handwriting datasets or use cases.

Model Used This project uses the microsoft/trocr-base-handwritten model, fine-tuned for handwritten text recognition.

Requirements Python 3.8 or later

Install dependencies:

bash Copy pip install -r requirements.txt Usage

1. Clone the Repository bash Copy git clone https://github.com/rafia9005/WriteVision.git cd WriteVision
2. Install Dependencies bash Copy pip install -r requirements.txt
3. Run the Program From URL To recognize text from an image URL:

python src/main.py --input url --source "https://fki.tic.heia-fr.ch/static/img/a01-122-02-00.jpg"

From Local Path To recognize text from a local image file:

bash Copy python src/main.py --input path --source "data/sample_image.jpg" From Base64 To recognize text from a Base64-encoded image string:

python src/main.py --input base64 --source "<base64-string>"

Replace with your actual Base64-encoded image data.

Example

From URL

python src/main.py --input url --source "https://fki.tic.heia-fr.ch/static/img/a01-122-02-00.jpg"

From Local Path

python src/main.py --input path --source "data/sample_image.jpg"

From Base64

python src/main.py --input base64 --source "iVBORw0KGgoAAAANSUhEUgAA..."

License This project is licensed under the MIT License. See LICENSE for details.

Feel free to contribute or adapt this project for your needs! 🚀