Using multi-task residual network to evaluate image aesthetic quality

With the popularity of cameras, smartphones and other devices, visual content data such as images and videos are increasing day by day, and visual content data processing technology has become a research direction in the research fields of computer vision, computational photography and artificial intelligence. Compared with the semantic features of object recognition, the mystery of human aesthetics has not yet been quantified and scientifically explained. It is difficult to determine the feature points of image aesthetics, which increases the difficulty of image aesthetics quality evaluation. In order to better obtain the feature points of image aesthetics, this paper uses a multi-task residual network model to extract features of the image, and evaluate the aesthetic level of the image according to the extracted features. Using a special aesthetic evaluation data set to test and verify the accuracy of the evaluation model proposed in this paper is 84.68%, which can more accurately evaluate the aesthetic level of images. Therefore, the evaluation model proposed in this paper promotes the development of image aesthetic value evaluation technology.