Leveraging big data in art education: Navigating the intelligent interaction journey with CNN algorithms for steering paths

This paper firstly investigates the interaction mode of art education and compares the effect of smart interaction with that of traditional classroom. The purpose of the research is using the weighting operation of the original feature map by the convolutional layer to extract the features of the image, integrating the features in the fully connected layer after several convolutional layers, and finally outputting the classification results. Then, the feature extraction backbone network DetNet-59 is constructed by using deformable convolution instead of traditional regular square convolution and combined with cavity convolution to make the model effectively increase the perceptual field. And the CNN intelligence algorithm is improved by adopting the guided region proposal network GA-RPN. Interaction detection performance of the proposed method outperforms all other methods; finally, comparison experiments and ablation experiments are conducted on the classical dataset V-COCO dataset and HICO-DET dataset to verify the effectiveness of the improved algorithm on the recognition of intelligent interaction behaviors in art education. On the V-COCO dataset, we achieved 62.5% mAP and a 3.7% mAP improvement, and the method in this paper outperforms other models by 10.5% mAP. Models utilizing CNN intelligence algorithms are notably more proficient in discerning and reasoning about intelligent interactions.