This paper analyzes and designs three problems existing in the new project curriculum of image classification tasks for software majors in the new engineering discipline. These problems include a lack of familiarity with Python language, a short time to learn image classification, and difficulty in setting up practical environments for students. To address these issues, a course teaching design based on Baidu’s AI Studio platform and centered around the application of pre-trained models from PaddleHub is proposed. This design incorporates practical applications of linear algebra, fine-tuning practices with pre-trained models, and online and offline team collaboration and learning. In the future, this design can be improved by allowing students to independently select and design recognition tasks. This curriculum design aims to help students lay a foundation in engineering, improve professional skills, cultivate teamwork abilities, and ultimately achieve the goal of enhancing professional interest and confidence.