Abstract:
Due to the large scale of registered trademark data, the similar mark search has become time-consuming and challenging. A similar mark search is an essential process that helps to avoid trademark infringement incidents. However similar mark search is a crucial process, there are major throwbacks in the existing workflow such as less accuracy, high time consumption and inefficiency. AutoTMvision uses a pre-trained Convolutional Neural Network (CNN) architecture to analyse a large set of trademark images and extract feature vectors for each of every trademark image. The main task will be that of "similar trademark search," which refers to searching for the most similar set of trademark images to some query trademark image. The results of the experiments reveal that the proposed method is substantially better and more accurate than standard trademark retrieval methods and is more efficient and timesaving.
