OCR && Document Search Web Application

• The "OCR Web Application with Keyword Search" prototype leverages the advanced capabilities of ColPali, Byaldi, and Huggingface Transformers for Qwen2-VL to perform Optical Character Recognition (OCR) on uploaded images containing both Hindi and English text. This web-based tool, developed using Gradio, offers a user-friendly interface for image upload, text extraction, and keyword search functionalities.

Key Features:

• Multilingual OCR: Utilizes the ColPali implementation of the Byaldi library in conjunction with Huggingface Transformers for Qwen2-VL, enabling accurate and efficient text extraction from images in Hindi and English.
• Keyword Search: Implements a basic search feature that allows users to locate specific keywords within the extracted text.
• Gradio Interface: Employs Gradio to build an interactive and intuitive web interface for seamless user interaction.
• Live Deployment: Accessible via a live URL, allowing users to experience the application's capabilities in real-time.

Technologies Used:

• Python: Programming language for backend development and model integration.
• ColPali: Efficient data processing and model training framework.
• Byaldi Library: Advanced natural language processing library.
• Huggingface Transformers: Pre-trained models and accelerated development tools.
• Qwen2-VL: Model specifically designed for visual language processing tasks.
• Gradio: Frontend framework for building interactive web applications.

How to Use:

• Upload an image containing both Hindi and English text to the web application.
• The ColPali-Byaldi-Huggingface Transformers model processes the image to extract text in both languages.
• View the extracted text displayed on the interface.
• Utilize the keyword search functionality to search for specific keywords within the extracted text.

Demo Link:

• https://drive.google.com/file/d/1WTBwLMgm-7kjjZbVLds6O-ox14ZfrxYW/view?usp=drivesdk