Fabrication Classifiation in Visual Tweets by Using Machine Learning with Deep Features
Vaishnavi J. Deshmukh, Asha Ambhaikar · 2024
In the context of online news outlets like Twitter, our work gives a summary of the situation as of rumor identification utilizing visual content. The majority of studies in the literature use visual content to illustrate several solutions to the issue. As opposed to combining them, we suggested an approach for Twitter rumours. The majority of our research work focuses on the use of supervised techniques, particularly deep learning models by utilising the deduced feature sets from the Twitter MediaEval dataset, which included 3000 images in total, a VGG-19 deep neural network was trained and evaluated. The VGG-19 achieves a highest sensitivity and an accuracy of using features taken from images during training and testing, which struggle to function in situations other than those in which they were trained. The simulation findings show that compared to other state-of-the-art architecture, the suggested design is more accurate and sensitive.