Research on Implementation Pathways for Computer Vision Technology in Dual-Track Teaching Reform of Innovation and Entrepreneurship Education for Vocational Higher Education Art Programs Driven by Artificial Intelligence

Abstract

This study integrates computer vision into the dual-track teaching reform of higher vocational art programs within the innovation and entrepreneurship framework. A reproducible analytical pipeline was developed using convolutional feature extraction, supervised classification, and ordinal regression, refined through enterprise calibration and domain-adaptive fine-tuning.Empirical analysis used a dataset of student artworks (N=4,320), curriculum metadata, and enterprise evaluations. Model performance was assessed via stratified cross-validation and bootstrap inference, with key metrics including MAE (5.9), accuracy (84.7%), F1-score, and curriculum alignment index. The model evaluated technical skill, creativity, and marketability based on an industry-developed rubric. Enterprise calibration improved competency alignment, and robustness was confirmed under challenging conditions.This research provides a scalable framework for automated visual assessment in dual-track curricula, offers guidance for university-enterprise collaboration, and highlights limitations in generalizability and interpretability. Recommendations include iterative recalibration, improved labeling, and longitudinal assessment to ensure pedagogical validity and ethical implementation.