Artwork Intelligent Recognition Bases On Mobile Terminal

Artwork analysis,detection and recognition is an important research direction in the field of image processing,which attracts a large number of researchers.However,most of the current proposed methods are not designed for the demand of real-time analysis with mobile devices.Moreover,existing methods usually rely on high quality images and require large amounts of computing consumption.With the development of deep learning,targeted analysis models have been widely applied in the fields of automatic driving and security monitoring.Artificial intelligence technology enables mobile devices to have the ability of machine learning,which provides a new direction for the development of artificial intelligence technology.Based on deep learning technology,this paper proposes a high-precision and lightweight artwork recognition framework via mobile devices.The innovations in this paper are as follows:1.A painting recognition system via mobile terminal.The system mainly consists of two parts,i.e.,painting detection module and painting recognition module.(1)The detection module adopts a new painting detection algorithm combined with painting landmark location.The landmark location module is added into the general detection algorithm to obtain the standard image for recognition.Two-stage detection method can effectively eliminate the influence of interference factors on recognition.(2)The application platform of the painting recognition system is the mobile terminal.Considering the limited computing capacity of the mobile devices,the recognition module adopts the MobileNet as the backbone.Moreover,the local features fusion operation great enhance the robustness of model.In order to validate the effectiveness of the proposed method,we have established two large scale artwork image databases.2.A Chinese seal recognition system based on ultra-light siamese network.The system mainly contains three units.(1)A new siamese network with multi-task learning,which can effectively solve the similarity measurement problem and improve the recognition accuracy of the model.(2)A new online data generation algorithm called Automatic Background Generation(ABG)which could generate numerous seal images with different backgrounds for effective training.(3)A new training method for siamese network which based on a central constraint idea.The training method is more scientific and effective and can improve the generalization of the model.In order to validate the effectiveness of the proposed method,we have established a large scale seal image databases.Experiments on this dataset show that our method is effective and advanced.