Research on an Evaluation Method for the Emotional Healing Effects of Abstract Color Field Art Based on Deep Convolutional Neural Networks

Abstract

Based on the theoretical foundations of art therapy and color effects, this paper proposes an abstract color space art emotional healing effect evaluation method using a deep multi-task convolutional neural network model. The proposed model consists of a dual-stream deep residual network, which is used for color feature extraction and multi-task learning training, respectively, to predict emotional healing effects and overall scores. Using the training set of 8,500 images from the AADB dataset as the research object, the results show that the accuracy rates predicted by the proposed model are 87.40% and 83.24%, respectively. Therefore, deep convolutional neural network models can be applied to the assessment of emotional healing effects. The three feature factors—color harmony, color vividness, and balance elements—all exhibit positive therapeutic effects. The therapeutic evaluation results for the three different groups are ranked as follows: Group II > Group I > Group III.