The old newspaper project

Abstract

In this project, we will create a newspaper image recognition app based on AI which uses Computer Vision, Optical Character Recognition (OCR) technology and Natural Language Processing (NLP) models to recognize, classify and extract information from images of newspapers. The project follows an iterative process; it begins with the investigation of Tesseract OCR and Yolov8 in a pre-alpha version to extract text and analyze layout. Having faced limitations we shifted to advanced implementations such as PaddleOCR along with LayoutParser for styling newspaper layouts and document parsing and using GPT-3 models for detailed outline writing of the extracted content., In this report, we describe the project context, technology choices, how we collected data is followed by some challenges we faced while building the application and how we solved those issues to improve the application's functionality.