Answers are grounded in this paper and cite their sources.
Controlled study demonstrates real-time neural style transfer guidance in oil painting students, indicating practical utility for personalized art education.
Oil-painting teaching faces persistent challenges in personalized creative guidance, limited technical feedback, and weak connection between skill training and aesthetic expression. To improve the computational support of individualized instruction, this study develops an AI-assisted creative guidance model based on neural style transfer. A VGG-19 convolutional neural network is used to extract multi-layer content and style features from oil-painting images, and Gram -matrix-based style representations are used to characterize brushstroke texture, color gradation, and contrast. A multistyle fusion algorithm is designed to combine multiple artistic styles through adaptive weighting and spatial partitioning, thereby reducing style conflict while maintaining controllable visual expression. A personalized teaching architecture is further established, including student profiling, style recommendation, real-time preview, and work-evaluation modules. A 16-week experiment with 68 oil-painting students compares the AI-assisted system with traditional instruction. Results show that the method achieves an IoU of 0.87 in style-region segmentation, a single-style inference time of 0.3 s, a preview generation time of 2.1 s, and sufficient GPU throughput for concurrent classroom use. The model provides a feasible framework for image signal processing, style-feature extraction, and intelligent art-education support.
No takes yet. Share an insight, caveat, or question.
W. Q. Li (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: