Visual Memory Neural Network for Artistic Graphic Design

Artistic graphic design is the aesthetic result of designer’s fusion various elements, with a high degree independence. Considering lack significant visual scope and indicators design, our research aims to build an upgraded network model that can categorize different types artistic graphics labels realize free combination solutions. We scheme reorganization from perspective computer vision propose method based on memory neural network. built environment reconstructed set up independent deep camera range calculation law. region segmentation problem, we self-attentive mechanism, which quantitatively segment regions according temporal features, before arranging them in sequence obtain feature vector. also add LSTM structure attention mechanism match self-attention features graphical module pass matched vector extract labeled text information graphs. To test effectiveness method, database adaptive training process. compared learning methods same type, experimental results proved outperformed other by keeping accuracy quantitative evaluation models above 90%.