|
Download PDFOpen PDF in browserOnline Fake Logo DetectionEasyChair Preprint 126656 pages•Date: March 21, 2024AbstractThe proliferation of digital media and the ease of content creation have given rise to a pressing issue – the spread of fake logos. Protecting the integrity of brand identities is crucial in the modern landscape, necessitating effective fake logo detection mechanisms. This research endeavors to address this challenge through the development of a robust detection system using Python and web browser URLs.The method- ology involves the acquisition of diverse datasets comprising authentic and manipulated logos, laying the foundation for a comprehensive training regimen. Employing convolutional neural networks CNN and leveraging deep learning frameworks like TensorFlow, the study aims to build a model capable of discerning subtle variations indicative of counterfeit logos. Pre- processing steps involve standardizing image sizes, normalizing pixel values, and augmenting data for model generalization. The model architecture incorporates convolutional layers for feature extraction and dense layers for classification, fostering the ability to distinguish between genuine and fabricated logos.To facilitate real-world application, the system utilizes web scraping techniques to extract logo images from web browser URLs. This integration enables the model to assess logos encountered in online environments, contributing to a proactive defense against logo-based misinformation.The implementation involves loading the trained model, pre processing web-scraped images, and utilizing the model for predictions. The model’s performance is evaluated based on its ability to accurately classify logos as authentic or fake. Keyphrases: 6. Deep learning, Flask, Image authenticity, Logo detection, Model Deployment, counterfeit detection, data preprocessing, feature extraction, file upload, image classification, image processing, machine learning, model evaluation, model training, web application Download PDFOpen PDF in browser |
|
|