In this paper, based on the convolutional neural network, ResNet recognition network and CBAM attention mechanism, fusing the advantages of the three algorithms, a deep convolutional block attention module (DL-CBAM) image recognition network is constructed. Using this model, a method based on the theme mining of digital resources of non-heritage skills is designed, and the model non-heritage digital resource mining performance as well as digital inheritance effect is analyzed through the constructed database of non-heritage skills. The model in this paper still has a better recognition effect on the cultural mechanism of Weinan’s non-heritage arts and crafts when the number of images is 200, and the accuracy rate is more than 90%. Specifically, it shows the superiority of the model in the identification of the emotional value of the non-legacy skills and culture for the classification of the characteristics of the non-legacy skills and culture. In addition, the non-heritage knowledge meta-extractability performance of the model in this paper has improved by 3.71% to 18.16% compared to the benchmark model. The DL-CBAM method improves students’ familiarity and satisfaction with non-heritage culture, and with an average rating of about 4.00 at the student behavioral level, the students are willing to borrow the model to promote the cultural digital inheritance. This study verifies the reliability of using deep learning models to identify the cultural value and digital inheritance of Weinan’s non-heritage arts and crafts in a university setting.