Identification of Fake Logo Detection Using Deep Learning
P. Vanitha, T Mohana Priya, P. Navasakthi, V. S. Rakshana Devi, R. Aarthi · 2024
The development and distribution of visual content, including logos, has increased as a result of the widespread use of digital media and online platforms. Fake logos, on the other hand, have also become more common as a result of this, and they can be used for misleading activities like disinformation or brand impersonation. It is essential to identify these phony logos to protect brand integrity and guarantee reliable visual data. Convolutional Neural Networks are utilized by the suggested system to automatically identify and extract pertinent features from logo images. The suggested approach focuses on making use of Efficient Net’s efficiency and scalability to accurately identify faint visual cues that indicate phony logos. To achieve accurate logo detection in an indoor environment, this work incorporates deep learning algorithms. Resource-constrained environments can benefit from a lightweight yet effective solution for logo detection thanks to MobileNet, which is wellknown for its effectiveness in real-time applications on mobile devices. The project entails building an extensive dataset with real and phony logo images that span a broad spectrum of styles, resolutions, and manipulation methodologies. The model’s accuracy in differentiating between real and fake logos is measured using performance metrics like F1 score, precision, and recall. To ensure robust model generalization, the dataset is carefully curated to include variations in scale, orientation, lighting conditions, and digital manipulations. In response to the growing problem of counterfeiting logos, this work presents a ResNet50-based method for identifying phony logos. Identify the logos by using these algorithms whether it is fake or real logos. By applying these different deep learning algorithms like CNN, ResNet50, MobileNet, and EfficientNet and compare the best algorithm based on high accuracy and efficiency.