Research on the construction of an art design teaching evaluation system combining CLIP and SF-IQA algorithms

Against the background of the digital transformation of art education, this study addresses two persistent problems in art design teaching evaluation: insufficient semantic evaluation of students' works and incomplete assessment of image quality. We construct a CLIP-SF-IQA (Contrastive Language-Image Pre-training-Spatial Feature Image Quality Assessment) evaluation system in which CLIP extracts image-text semantic features related to theme, style, and design intention, while SF-IQA estimates no-reference technical quality; an adaptive feature-level fusion module then produces the comprehensive evaluation result. The fusion system achieves color coordination, composition logic, and style consistency matching degrees of 89.7%, 87.4%, and 91.2%, respectively. Its Pearson correlation with the aggregated teacher score is 0.91 with a mean absolute error of 3.2 points, and its average evaluation time is 2.1 min per work, corresponding to a 5.71-fold speed-up over manual evaluation. These results indicate that the proposed system can provide efficient and interpretable auxiliary evidence for preliminary assignment evaluation, staged feedback, and work diagnosis, while final pedagogical judgments remain under teacher supervision.