This paper takes CDGAN as the core technology framework, combines the data-user collaboration model with the reconstruction of the curriculum resource system, and explores the collaborative innovation path of visual communication design and interaction design education in the intelligent era. By introducing DO-Conv and CA, we optimize the number of parameters and feature extraction ability of CycleGAN image generation model, and realize the significant improvement in detail performance and semantic consistency of the generated images. Compared with the original model CycleGAN, the Loss value of CDGAN model is reduced by 0.016, the BLEU score is improved by 5.9%, and the performance is optimal in both FID and 1-NNA in both category and company datasets. The CDGAN model image clarity, vividness, and harmony index scores are 3.297, 3.286, and 3.278, respectively, which all have significant advantages over other methods. In the teaching experiment, there is a significant difference between the pre-test data and post-test data of the experimental group in the three dimensions of design thinking, such as variability, uniqueness and delicacy, and the cultivation of design thinking ability (P<0.05), and the post-test scores of each dimension are about 2 points higher than the pre-test scores, and the design thinking ability reaches 87.42±4.286 points. This paper constructs a technology-enabled teaching model to provide theoretical basis and practical reference for the transformation of visual communication design education.
