Multimodal Misinformation Detection

Project Code :TCMAPY1991

Objective

The Multimodal Misinformation Detection system detects misleading content by analyzing both text and images using advanced AI models. For text analysis, transformer-based models like BERT, RoBERTa, and XLNet are used to capture contextual meaning, while for image verification, Convolutional Neural Networks (CNN) and MobileNet are employed to identify manipulated or authentic images. The system integrates Explainable AI techniques such as SHAP to ensure transparency in predictions and is implemented as a Flask web application. This comprehensive, scalable framework enables accurate, interpretable, and automated detection of misinformation across multiple modalities, enhancing reliability in digital content verification.

Abstract

Misinformation in both textual and visual content has become a significant concern in information analysis. This project, Multimodal Misinformation Detection, focuses on identifying inaccurate or misleading content by analyzing both news text and images. For text classification, transformer-based models such as BERT, RoBERTa, and XLNet are used, alongside CNN and FastText models, providing high accuracy in detecting false news articles. Explainable AI techniques, including SHAP, are integrated to visualize and understand model predictions, making the process interpretable. For image verification, convolutional neural networks and MobileNet are utilized to distinguish manipulated images from authentic ones. The system is implemented as a web application using Flask, with modules for registration, login, news classification, and logout. The approach combines natural language processing and computer vision, enabling reliable detection of misinformation across multiple modalities. The project demonstrates an effective framework for automated detection, offering insights into content verification and model decision-making.

Keywords: Misinformation detection, Fake news, Text classification, Image classification, BERT, RoBERTa, CNN, MobileNet, Explainable AI, SHAP

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.