Research on Multidimensional Evaluation Method of Digital Art Works Based on Multimodal Data Fusion Technology

Abstract

Aiming at the problems that the traditional evaluation model does not realize the full use of information, resulting in poor generalization ability and low prediction accuracy, this paper combines the BPNN algorithm with the Self-attention mechanism, and puts forward a multi-dimensional evaluation model for digital art works based on SA-BPNN multimodal fusion. A multi-dimensional evaluation index system for digital art works is constructed, in which the weights of the indexes are obtained through the hierarchical analysis method. It is found that the aesthetic value is the most important in the final rating of digital art works, and emotional resonance and creative autonomy are also important evaluation elements. The study validated the weights of the AHP method by constructing a back-propagation neural network, and after validation, the prediction accuracy of the evaluation model reached 98.85%, which proved the effectiveness of the evaluation model. This paper provides theoretical support and tools for the realization of multi-dimensional evaluation of digital art works by using multimodal data fusion technology.