Image based cyberbullying detection using deep learning

Cyberbullying is a growing issue in the digital world, especially on social media platforms where users frequently upload images. The rise in online harassment and its impact on individuals, particularly adolescents, necessitates effective tools for detecting harmful content in images. This study proposes a deep learning-based approach for image-based cyberbullying detection using Convolutional Neural Networks (CNN). Specifically, the research leverages the VGG-16 and MobileNet architectures, both of which are renowned for their efficiency in image classification tasks. In the proposed model, the input consists of images from social media platforms, which are processed to identify subtle signs of cyberbullying. VGG-16, a deep CNN architecture known for its ability to extract hierarchical features, is employed for its strong performance in image classification tasks. MobileNet, on the other hand, is utilized for its lightweight design, making it suitable for real-time applications on mobile devices while maintaining high accuracy. The system works by training these models on a dataset of labeled images containing cyberbullying and non-cyberbullying content. The models are evaluated based on accuracy, precision, recall, and F1-score to measure their effectiveness in identifying harmful content. The proposed solution aims to provide an automated, scalable, and efficient method to detect cyberbullying in images, ultimately helping to create safer digital environments by identifying and flagging harmful content in real-time. Keywords: Cyberbullying Detection, Deep Learning, Image Classification, Convolutional Neural Networks, VGG-16, MobileNet, Social Media, Real-Time Detection.