A NOVEL IMAGE-AND-TEXT-PLAGIARISM-DETECTION

Abstract

Plagiarism has become a major issue in the academic and digital world, especially with the increasing availability of online documents, articles, and media. Today, individuals can easily copy text and images from the internet, modify them, and present them as original work, leading to copyright violations and ethical concerns. Traditional plagiarism detection systems mainly focus on text-only similarity and fail to detect plagiarism involving visual content such as images. Therefore, there is a need for a modern plagiarism detection system capable of analyzing both text and images. This project proposes a combined text and image plagiarism detection system designed to identify copied or similar content across documents and digital images. For text plagiarism detection, methods such as Natural Language Processing, tokenization, semantic similarity, and cosine similarity are used to detect exact or paraphrased plagiarism even after changing sentence structure or wording. These techniques allow deeper understanding and comparison of meaning rather than depending on direct text matching. For image plagiarism detection, the system applies computer vision and deep learning techniques to identify similarities between images. Feature extraction models such as ORB, CNN and SSIM enable detection of altered, rotated, resized, filtered, or partially modified images.