Art teaching interaction based on multimodal information fusion under the background of deep learning

The advancement of information technology is crucial to the reform of teaching, and the requirements for interaction and convenience in art teaching are getting higher. The creation of an interactive art platform for art teaching requires the collection and classification of a large number of art works. However, enough human resources are required to classify a high volume of art works. Similarly, the deep learning technology can automatically classify images, but the complexity and abstraction of art composition make it difficult to refine the classification. This paper presents an automatic classification method of art teaching works based on improved deep learning model resnet101 and multimodal information fusion. We take the text information related to the description of art works as the features of another dimension and use bottleneck module to improve the original model, so as to improve the accuracy of model classification. Experiments show that the method in this paper improves sensitivity, specificity, precision, F1 score and accuracy by 8.82%, 10.21%, 7.84% and 9.41%, respectively. The proposed method effectively facilitates the construction of art teaching platform and makes contributions to art interactive teaching.

Automated summary

The advancement of information technology is crucial to the reform of teaching, and the requirements for interaction and convenience in art teaching are getting higher. This paper presents an automatic classification method of art teaching works based on improved deep learning model resnet101 and multimodal information fusion. The proposed method effectively facilitates the construction of art teaching platform and makes contributions to art interactive teaching.