According to the constructivist learning theory, learners need to actively construct a knowledge system in a practical situation, and deep learning, with intelligent data analysis, pattern recognition and predictive decision-making capabilities, just builds a practical platform for students to discover market opportunities and formulate innovative solutions, which effectively promotes the development of their innovation and entrepreneurship skills. At the same time, deep learning emphasizes the understanding, utilization and generation of knowledge, and further strengthens students’ innovation and entrepreneurship literacy by cultivating human-machine collaboration, which is in line with the concept of “learning by doing” in contextual cognitive theory. In addition, deep learning has the characteristics of thinking training, knowledge transfer and application, which echoes the requirements of higher-order thinking cultivation in Bloom’s educational goal classification theory, which can effectively improve the knowledge integration, transformation and application ability of college students, and enhance the ability of knowledge creation and innovation and entrepreneurship. In the application of practical education scenarios, taking the YOLOv5s network as an example, after the