StyleBusters: Protecting the intellectual property rights of designers using deep learning techniques

Intellectual property rights are the legal rights that protect the creations of an owner's mind, such as inventions, artistic works, and designs. Designers play a crucial role in the creative industry, and intellectual property rights ensure that their designs are protected from infringement by others. As such, designers can control how their designs are used and monetized. In this paper, we have introduced a styles visual recommendation system that detects the similar style designs in three professional domains: artworks, clothes, and logos. The VGG (visual Geometry Group) classifier is trained to extract style features for the artworks and clothes domains, and a pre-trained MobileNet model is used as a feature extractor for the logos domain. YOLO V8 object detection model is trained for the clothes domain. Web scraping techniques are relied on to search for and retrieve artworks, clothes, and logos from the most reliable and largest websites, which saves storage and maintenance costs. Then we detect similar style artworks, clothes, and logos based on the cosine similarity between the features embedded in the input images and those retrieved from the websites. The achieved performance metrics are as follows: validation accuracy equals 77.0377% using the VGG model on the WikiArt Dataset, box-loss equals 0.3342, and a classification-loss equals 0.5302 for the object detection of 13 different clothing categories using YOLO-V8. A validation accuracy of 65% and a training accuracy of 68% for classifying 46 clothing categories using the VGG-ll model was obtained.