Aesthetic Attribute Assessment of Images Numerically on Mixed Multi-attribute Datasets

With the continuous development of social software and multimedia technology, images have become a kind important carrier for spreading information socializing. How to evaluate an image comprehensively has focus recent researches. The traditional aesthetic assessment methods often adopt single numerical overall scores, which certain subjectivity can no longer meet higher requirements. In this article, we construct new attribute dataset called mixed with attributes (AMD-A) design external features fusion. Besides, propose efficient method on multi-attribute multitasking network architecture by using EfficientNet-B0 as backbone network. Our model achieve classification, scoring, scoring. each sub-network, improve feature extraction through ECA channel attention module. As final idea teacher-student use classification sub-network guide fine-grain regression. Experimental results, MindSpore, show that our proposed effectively performance assessment.