Abstract: Conventional image post-processing procedures in photomechanics experiments are complex and require specialized knowledge and data-processing skills, making them difficult for students to master quickly and delaying feedback on experimental results. To address these issues, this study proposes a teaching reform approach that integrates artificial intelligence (AI) into photomechanics laboratory teaching. Taking photoelasticity, digital image correlation (DIC), and digital gradient sensing (DGS) as representative techniques, a convolutional neural network based on U–Net is employed to achieve unified identification and rapid analysis of experimental images. In addition, AI–assisted visualization and analysis software is developed for photomechanics experiments to transform complex image processing, model inference, and field reconstruction procedures into intuitive and accessible interactive operations. This enables students to obtain quantitative experimental results promptly during class and to analyze mechanical phenomena and discuss relevant theories based on the visualized results. The proposed approach reduces the complexity of conventional photomechanics data processing, improves the timeliness and interactivity of laboratory teaching, and promotes inquiry-oriented learning, thereby providing a practical pathway for AI–assisted reform of experimental mechanics education.