This study proposes a novel DMA-CNN framework for enhancing teaching methods in digital media art by leveraging the power of convolutional neural networks. Unlike existing approaches such as Creative Intelligence Cloud, DL-ALS, DCNN-Shallow NN, GAN surrogate, and a CNN baseline, the proposed model is designed to integrate feature-aware evaluation, immersive learning, and adaptive feedback, thereby fostering creativity and personalization in art education. The key contribution of this work lies in developing a specialized CNN architecture tailored for evaluating and guiding digital art learning and teaching, and benchmarking it against multiple baselines on a curated dataset. Experimental results demonstrate that DMA-CNN outperforms all other models, achieving 96% accuracy, 96% precision, 96% recall, 95.9% F1-score, and 0.998 ROC-AUC, confirming its effectiveness and scalability for advancing digital media art pedagogy.
Wang et al. (2026) studied this question.